Model tuning for cross-node machine learning

Abstract

Methods, systems, and devices for wireless communication are described. A user equipment (UE) may obtain data samples of a first machine learning model associated with a task at the UE. A first set of parameters may be associated with the first machine learning model. The UE may send a capability message indicating a capability of the UE to perform a tuning process of the first machine learning model, and the UE may perform the tuning process of the first machine learning model based on the capability of the UE to perform the tuning process of the first machine learning model, to obtain a second set of parameters associated with the first machine learning model. The capability by which the UE performs the tuning procedure may be one of an online tuning capability or an offline tuning capability.

Description

Technical Field

[0001] The following relates to wireless communication, including model tuning for cross-node machine learning. Background Art

[0002] Wireless communication systems are widely deployed to provide various types of communication content, such as voice, video, packet data, messaging, broadcasting, and so on. These systems may be capable of supporting communication with multiple users by sharing available system resources (e.g., time, frequency, and power). Examples of such multi-access systems include fourth-generation (4G) systems (such as Long-Term Evolution (LTE) systems, LTE-Advanced (LTE-A) systems, or LTE-A Pro systems) and fifth-generation (5G) systems (which may be referred to as New Radio (NR) systems). These systems may employ techniques such as Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal FDMA (OFDMA), or Discrete Fourier Transform Spread Orthogonal Frequency Division Multiplexing (DFT-S-OFDM). A wireless multi-access communication system may include one or more base stations, each of which supports wireless communication for communication devices, which may be referred to as User Equipment (UE). Summary of the Invention

[0003] The described techniques relate to improved methods, systems, devices, and apparatuses for supporting model tuning for cross-node machine learning. Generally speaking, the described techniques enable a User Equipment (UE) to autonomously perform a tuning (e.g., fine-tuning) process of a machine learning model used in communication between the UE and a network entity. For example, the UE may receive data samples (e.g., a training data set) from the network entity to train the machine learning model. The UE may send a capability message to the network entity, and the capability message may indicate whether the UE can autonomously perform the tuning process. The UE may generate a second training data set based on the tuning process, and the UE or the network entity may use the second training data set to perform a tuning process of a second machine learning model.

[0004] A method for wireless communication at a UE is described. The method may include: obtaining data samples of a first machine learning model associated with a task at the UE, wherein a first set of parameters is associated with the first machine learning model; sending a capability message indicating the capability of the UE to perform a tuning process of the first machine learning model; performing the tuning process of the first machine learning model based on the capability of the UE to perform the tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model; and sending a message indicating at least a portion of the second set of parameters to a network entity based on performing the tuning process of the first machine learning model.

[0005] Describes an apparatus for wireless communication at a UE. The apparatus may include: a processor; a memory coupled to the processor; and instructions stored in the memory. The instructions may be executable by the processor to cause the apparatus to perform the following operations: obtain a data sample of a first machine learning model associated with a task at the UE, wherein a first set of parameters is associated with the first machine learning model; send a capability message indicating the UE's ability to perform a tuning process of the first machine learning model; perform the tuning process of the first machine learning model based on the UE's ability to perform the tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model; and send a message indicating at least a portion of the second set of parameters to a network entity based on performing the tuning process of the first machine learning model.

[0006] Describes another apparatus for wireless communication at a UE. The apparatus may include: means for obtaining a data sample of a first machine learning model associated with a task at the UE, wherein a first set of parameters is associated with the first machine learning model; means for sending a capability message indicating the UE's ability to perform a tuning process of the first machine learning model; means for performing the tuning process of the first machine learning model based on the UE's ability to perform the tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model; and means for sending a message indicating at least a portion of the second set of parameters to a network entity based on performing the tuning process of the first machine learning model.

[0007] Describes a non-transitory computer-readable medium storing code for wireless communication at a UE. The code may include instructions executable by a processor to perform the following operations: obtain a data sample of a first machine learning model associated with a task at the UE, wherein a first set of parameters is associated with the first machine learning model; send a capability message indicating the UE's ability to perform a tuning process of the first machine learning model; perform the tuning process of the first machine learning model based on the UE's ability to perform the tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model; and send a message indicating at least a portion of the second set of parameters to a network entity based on performing the tuning process of the first machine learning model.

[0008] In some examples of the methods, apparatuses, and non-transitory computer-readable media described herein, the first machine learning model includes an encoder portion of a second machine learning model, and a third machine learning model includes a decoder portion of the second machine learning model.

[0009] In some examples of the methods, apparatuses, and non-transitory computer-readable media described herein, performing the tuning process of the first machine learning model may include operations, features, components, or instructions for performing the following actions: receiving a set of parameters associated with a loss function.

[0010] Some examples of the methods, apparatuses, and non-transitory computer-readable media described herein may further include operations, features, components, or instructions for performing the following actions: sending a message associated with a forward propagation process.

[0011] Some examples of the methods, apparatuses, and non-transitory computer-readable media described herein may further include operations, features, components, or instructions for performing the following actions: receiving a message associated with a backpropagation process for adjusting parameters associated with an encoder, where the message indicates a gradient associated with the loss function; and updating the parameters associated with the encoder based on the message.

[0012] Some examples of the methods, apparatuses, and non-transitory computer-readable media described herein may further include operations, features, components, or instructions for performing the following actions: receiving a set of parameters associated with the loss function; and updating parameters associated with a decoder portion of the second machine learning model based on the set of parameters.

[0013] In some examples of the methods, apparatuses, and non-transitory computer-readable media described herein, sending the capability message may include operations, features, components, or instructions for performing the following actions: sending an indication of a set of machine learning models supported by the UE.

[0014] In some examples of the methods, apparatuses, and non-transitory computer-readable media described herein, the task includes a channel state information (CSI) feedback task.

[0015] In some examples of the methods, apparatuses, and non-transitory computer-readable media described herein, performing the tuning process of the first machine learning model may include operations, features, components, or instructions for performing the following actions: using the second set of parameters to update parameters associated with an encoder, parameters associated with a decoder, or both.

[0016] In some examples of the methods, apparatuses, and non-transitory computer-readable media described herein, performing the tuning process of the first machine learning model may include operations, features, components, or instructions for performing the following actions: performing an online tuning process.

[0017] In some examples of the methods, apparatuses, and non-transitory computer-readable media described herein, performing the online tuning process of the first machine learning model may include operations, features, components, or instructions for performing the following actions: when performing the task, updating the second parameter set associated with the encoder using the second parameter set for the first machine learning model.

[0018] In some examples of the methods, apparatuses, and non-transitory computer-readable media described herein, performing the tuning process of the first machine learning model may include operations, features, components, or instructions for performing the following actions: performing an offline tuning process.

[0019] In some examples of the methods, apparatuses, and non-transitory computer-readable media described herein, performing the offline tuning process of the first machine learning model may include operations, features, components, or instructions for performing the following actions: when performing the task, updating the second parameter set associated with the encoder using the first parameter set for the first machine learning model.

[0020] Some examples of the methods, apparatuses, and non-transitory computer-readable media described herein may also include operations, features, components, or instructions for performing the following actions: sending a third indication to the network entity, where the third indication indicates the availability of a second encoder, and the second encoder may be associated with performing the offline tuning process.

[0021] In some examples of the methods, apparatuses, and non-transitory computer-readable media described herein, performing the tuning process may include operations, features, components, or instructions for performing the following actions: receiving a first indication from the network entity associated with performing the tuning process of the first machine learning model, where the first indication includes an activation state or an allowed state; and in response to the first indication, sending a second indication to the network entity, where the second indication includes an activation indication associated with starting to perform the tuning process or a deactivation indication associated with stopping the tuning process.

[0022] Some examples of the methods, apparatuses, and non-transitory computer-readable media described herein may also include operations, features, components, or instructions for performing the following actions: sending an activation request to the network entity, where receiving the first indication may be based on the activation request.

[0023] Some examples of the methods, apparatuses, and non-transitory computer-readable media described herein may also include operations, features, components, or instructions for performing the following actions: receiving a fourth indication from the network entity, where the fourth indication includes a deactivation indication associated with stopping the tuning process.

[0024] In some examples of the methods, apparatuses, and non-transitory computer-readable media described herein, performing the tuning process of the first machine learning model may include operations, features, components, or instructions for performing the following actions: updating the parameters associated with the encoder, the parameters associated with the decoder, or both, using the second parameter set, based on the gradients associated with the first parameter set and the second parameter set.

[0025] In some examples of the methods, apparatuses, and non-transitory computer-readable media described herein, updating the parameters associated with the encoder, the parameters associated with the decoder, or both may include operations, features, components, or instructions for performing the following actions: receiving a parameter set associated with a loss function; and sending a message associated with a forward propagation process.

[0026] Some examples of the methods, apparatuses, and non-transitory computer-readable media described herein may further include operations, features, components, or instructions for performing the following actions: receiving a message associated with a backpropagation process for adjusting the parameters associated with the encoder, the parameters associated with the decoder, or both, where the message indicates a gradient associated with the loss function; and updating the parameters associated with the encoder, the parameters associated with the decoder, or both, based on the message.

[0027] In some examples of the methods, apparatuses, and non-transitory computer-readable media described herein, the UE receives the indication of the first parameter set via broadcast signaling, dedicated signaling, or both.

[0028] A method for wireless communication at a UE using a first parameter set associated with a first machine learning model for a task is described. The method may include: sending a capability message indicating the UE's capability to perform a tuning process of the first machine learning model to a network entity, where the first parameter set is associated with the first machine learning model; performing the tuning process of the first machine learning model based on the UE's capability to perform the tuning process of the first machine learning model to obtain a second parameter set associated with the first machine learning model; sending a message indicating the second parameter set to the network entity based on performing the tuning process of the first machine learning model; receiving an allowed status indication from the network entity associated with the second parameter set; performing the task using the first parameter set associated with the first machine learning model or the second parameter set associated with the first machine learning model based on the received allowed status indication; receiving a not-allowed status indication from the network entity associated with the second parameter set; and performing the task using the first parameter set associated with the first machine learning model based on the received not-allowed status indication.

[0029] A device for wireless communication at a UE using a first set of parameters associated with a first machine learning model for a task is described. The device may include: a processor; a memory coupled to the processor; and instructions stored in the memory. The instructions may be executable by the processor to cause the device to: send a capability message to a network entity indicating the UE's ability to perform a tuning process of the first machine learning model, where the first set of parameters is associated with the first machine learning model; perform the tuning process of the first machine learning model based on the UE's ability to perform the tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model; send a message indicating the second set of parameters to the network entity based on performing the tuning process of the first machine learning model; receive an allowed status indication from the network entity associated with the second set of parameters; perform the task using the first set of parameters associated with the first machine learning model or the second set of parameters associated with the first machine learning model based on the received allowed status indication; receive a not-allowed status indication from the network entity associated with the second set of parameters; and perform the task using the first set of parameters associated with the first machine learning model based on the received not-allowed status indication.

[0030] Another device for wireless communication at a UE using a first set of parameters associated with a first machine learning model for a task is described. The device may include: means for sending a capability message to a network entity indicating the UE's ability to perform a tuning process of the first machine learning model, where the first set of parameters is associated with the first machine learning model; means for performing the tuning process of the first machine learning model based on the UE's ability to perform the tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model; means for sending a message indicating the second set of parameters to the network entity based on performing the tuning process of the first machine learning model; means for receiving an allowed status indication from the network entity associated with the second set of parameters; means for performing the task using the first set of parameters associated with the first machine learning model or the second set of parameters associated with the first machine learning model based on the received allowed status indication; means for receiving a not-allowed status indication from the network entity associated with the second set of parameters; and means for performing the task using the first set of parameters associated with the first machine learning model based on the received not-allowed status indication.

[0031] Describes a non-transitory computer-readable medium that stores code for wireless communication at a UE using a first set of parameters associated with a first machine learning model for a task. The code may include instructions executable by a processor to perform the following operations: send a capability message to a network entity indicating the UE's ability to perform a tuning process of the first machine learning model, where the first set of parameters is associated with the first machine learning model; perform the tuning process of the first machine learning model based on the UE's ability to perform the tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model; send a message indicating the second set of parameters to the network entity based on performing the tuning process of the first machine learning model; receive an allowed status indication from the network entity associated with the second set of parameters; perform the task using the first set of parameters associated with the first machine learning model or the second set of parameters associated with the first machine learning model based on the received allowed status indication; receive a not-allowed status indication from the network entity associated with the second set of parameters; and perform the task using the first set of parameters associated with the first machine learning model based on the received not-allowed status indication.

[0032] Some examples of the methods, apparatuses, and non-transitory computer-readable media described herein may also include operations, features, components, or instructions for performing the following actions: autonomously determine whether to use the first set of parameters associated with the first machine learning model or the second set of parameters associated with the first machine learning model to perform the task, at least in part based on the received allowed status indication.

[0033] Describes a method for wireless communication at a UE using a first set of parameters associated with a first machine learning model for a task. The method may include: sending a capability message to a network entity indicating the UE's ability to perform an online tuning process of the first machine learning model, where the first set of parameters is associated with the first machine learning model; receiving a first indication from the network entity associated with the online tuning process, where the first indication includes an activation status or an allowed status; and performing the tuning process of the first machine learning model based on the UE's ability to perform the online tuning process of the first machine learning model and the received allowed status to obtain a second set of parameters associated with the first machine learning model.

[0034] A device for wireless communication at a UE using a first set of parameters associated with a first machine learning model for a task is described. The device may include: a processor; a memory coupled to the processor; and instructions stored in the memory. The instructions may be executable by the processor to cause the device to perform the following operations: send a capability message to a network entity indicating the UE's ability to perform an online tuning process of the first machine learning model, wherein the first set of parameters is associated with the first machine learning model; receive a first indication from the network entity associated with the online tuning process, wherein the first indication includes an activation state or an allowed state; and perform the tuning process of the first machine learning model based on the UE's ability to perform the online tuning process of the first machine learning model and the received allowed state to obtain a second set of parameters associated with the first machine learning model.

[0035] Another device for wireless communication at a UE using a first set of parameters associated with a first machine learning model for a task is described. The device may include: means for sending a capability message to a network entity indicating the UE's ability to perform an online tuning process of the first machine learning model, wherein the first set of parameters is associated with the first machine learning model; means for receiving a first indication from the network entity associated with the online tuning process, wherein the first indication includes an activation state or an allowed state; and means for performing the tuning process of the first machine learning model based on the UE's ability to perform the online tuning process of the first machine learning model and the received allowed state to obtain a second set of parameters associated with the first machine learning model.

[0036] A non-transitory computer-readable medium storing code for wireless communication at a UE using a first set of parameters associated with a first machine learning model for a task is described. The code may include instructions executable by a processor to perform the following operations: send a capability message to a network entity indicating the UE's ability to perform an online tuning process of the first machine learning model, wherein the first set of parameters is associated with the first machine learning model; receive a first indication from the network entity associated with the online tuning process, wherein the first indication includes an activation state or an allowed state; and perform the tuning process of the first machine learning model based on the UE's ability to perform the online tuning process of the first machine learning model and the received allowed state to obtain a second set of parameters associated with the first machine learning model.

[0037] Describes a method for wireless communication at a network entity. The method may include: receiving, from a UE, a capability message indicating the UE's capability to perform a tuning process of a first machine learning model associated with a first set of parameters at the UE; and receiving, from the UE, a message indicating at least a portion of a second set of parameters.

[0038] Describes an apparatus for wireless communication at a network entity. The apparatus may include: a processor; a memory coupled to the processor; and instructions stored in the memory. The instructions may be executable by the processor to cause the apparatus to perform the following operations: receiving, from a UE, a capability message indicating the UE's capability to perform a tuning process of a first machine learning model associated with a first set of parameters at the UE; and receiving, from the UE, a message indicating at least a portion of a second set of parameters.

[0039] Describes another apparatus for wireless communication at a network entity. The apparatus may include: means for receiving, from a UE, a capability message indicating the UE's capability to perform a tuning process of a first machine learning model associated with a first set of parameters at the UE; and means for receiving, from the UE, a message indicating at least a portion of a second set of parameters.

[0040] Describes a non-transitory computer-readable medium storing code for wireless communication at a network entity. The code may include instructions executable by a processor to perform the following operations: receiving, from a UE, a capability message indicating the UE's capability to perform a tuning process of a first machine learning model associated with a first set of parameters at the UE; and receiving, from the UE, a message indicating at least a portion of a second set of parameters.

[0041] In some examples of the methods, apparatuses, and non-transitory computer-readable media described herein, the first machine learning model includes an encoder portion of a second machine learning model, and a third machine learning model includes a decoder portion of the second machine learning model.

[0042] In some examples of the methods, apparatuses, and non-transitory computer-readable media described herein, receiving the capability message may include operations, features, means, or instructions for performing the following action: receiving an indication of a set of machine learning models supported by the UE.

[0043] Some examples of the methods, apparatuses, and non-transitory computer-readable media described herein may further include operations, features, means, or instructions for performing the following action: receiving the message may be associated with receiving channel state information feedback.

[0044] Some examples of the methods, apparatuses, and non-transitory computer-readable media described herein may also include operations, features, components, or instructions for performing the following actions: sending a first indication associated with performing a tuning process of the first machine learning model to the UE, where the first indication includes an activation state or a permission state; and receiving, in response to the first indication, a second indication from the UE, where the second indication includes an activation indication associated with starting to perform the tuning process or a deactivation indication associated with stopping the tuning process.

[0045] Some examples of the methods, apparatuses, and non-transitory computer-readable media described herein may also include operations, features, components, or instructions for performing the following actions: receiving an activation request from the UE, where sending the first indication may be based on the activation request.

[0046] Some examples of the methods, apparatuses, and non-transitory computer-readable media described herein may also include operations, features, components, or instructions for performing the following actions: sending a third indication to the UE, where the third indication includes a deactivation indication associated with stopping the tuning process. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 Examples of wireless communication systems that support model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure are illustrated.

[0048] Figure 2 Examples of wireless communication systems that support model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure are illustrated.

[0049] Figure 3 Examples of process flows that support model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure are illustrated.

[0050] Figure 4 Examples of process flows that support model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure are illustrated.

[0051] Figure 5 Examples of process flows that support model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure are illustrated.

[0052] Figure 6 Examples of process flows that support model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure are illustrated.

[0053] Figure 7 Examples of process flows that support model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure are illustrated.

[0054] Figure 8 Illustrates an example of a process flow that supports model tuning for cross-node machine learning according to one or more aspects of the present disclosure.

[0055] Figure 9 And Figure 10 Illustrates a block diagram of a device that supports model tuning for cross-node machine learning according to one or more aspects of the present disclosure.

[0056] Figure 11 Illustrates a block diagram of a communication manager that supports model tuning for cross-node machine learning according to one or more aspects of the present disclosure.

[0057] Figure 12 Illustrates a diagram of a system that includes a device that supports model tuning for cross-node machine learning according to one or more aspects of the present disclosure.

[0058] Figure 13 And Figure 14 Illustrates a block diagram of a device that supports model tuning for cross-node machine learning according to one or more aspects of the present disclosure.

[0059] Figure 15 Illustrates a block diagram of a communication manager that supports model tuning for cross-node machine learning according to one or more aspects of the present disclosure.

[0060] Figure 16 Illustrates a diagram of a system that includes a device that supports model tuning for cross-node machine learning according to one or more aspects of the present disclosure.

[0061] Figures 17 to 21 Illustrates a flowchart of a method that shows support for model tuning for cross-node machine learning according to one or more aspects of the present disclosure. Detailed Description

[0062] In some wireless communication networks, a user equipment (UE) may use data samples (e.g., a training data set) to train a machine learning model for communicating with the UE and network entities. In some examples, the UE may perform a tuning (e.g., fine-tuning) process of the training data set to improve the performance of the machine learning model and the performance of an encoder (e.g., at the UE) and a decoder (e.g., at the network entity). For example, the UE may have a baseline machine learning model (e.g., a first machine learning model), and the network entity may enable or direct the UE to perform a tuning process of the first machine learning model using the training data set via signaling. However, the network entity sending signaling to trigger the tuning process at the UE may result in increased latency and overhead.

[0063] The techniques, systems, and devices of this disclosure enable a UE to perform model tuning for cross-node machine learning. Generally speaking, the described techniques enable the UE to autonomously perform a tuning (e.g., fine-tuning) process for a machine learning model used in communication between the UE and a network entity. For example, the UE may send a capabilities message to the network entity, which indicates whether the UE can autonomously perform the tuning process. The tuning process can be either an online tuning process or an offline tuning process. The UE may generate a second training dataset based on the tuning process, and the UE or the network entity may use the second training dataset to perform a tuning process for a second machine learning model for use at a corresponding encoder, decoder, or both. The UE may generate the second training dataset according to a reference signal received from the network entity.

[0064] Aspects of the present disclosure are first described in the context of a wireless communication system. Aspects of the present disclosure are further illustrated by process flows. Aspects of the present disclosure are further illustrated and described by and with reference to apparatus diagrams, system diagrams, and flowcharts related to model tuning for cross-node machine learning.

[0065] Figure 1 An example of a wireless communication system 100 that supports model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure is illustrated. The wireless communication system 100 may include one or more network entities 105, one or more UEs 115, and a core network 130. In some examples, the wireless communication system 100 may be a Long Term Evolution (LTE) network, an Advanced LTE (LTE-A) network, an LTE-A Pro network, a New Radio (NR) network, an Advanced 5G network, or a network operating according to other system and radio technologies (including future system and radio technologies not explicitly mentioned herein).

[0066] The network entities 105 may be dispersed throughout a geographic area to form the wireless communication system 100 and may include devices in different forms or with different capabilities. In various examples, the network entities 105 may be referred to as network elements, mobility elements, radio access network (RAN) nodes, or network equipment, etc. In some examples, the network entities 105 and the UEs 115 may communicate wirelessly via one or more communication links 125 (e.g., radio frequency (RF) access links). For example, the network entity 105 may support a coverage area 110 (e.g., a geographic coverage area) within which the UEs 115 and the network entity 105 may establish one or more communication links 125. The coverage area 110 may be an example of a geographic area within which the network entity 105 and the UEs 115 may support signal communication according to one or more radio access technologies (RATs).

[0067] UE 115 can be dispersed throughout the coverage area 110 of the wireless communication system 100, and each UE 115 can be stationary or mobile or stationary and mobile at different times. The UE 115 can be a device in different forms or with different capabilities. Figure 1 Some example UEs 115 are illustrated therein. The UEs 115 described herein may be capable of supporting communication with various types of devices (such as other UEs 115 or network entities 105 as Figure 1 shown).

[0068] As described herein, a node of the wireless communication system 100 (which may be referred to as a network node or a wireless node) can be a network entity 105 (e.g., any network entity described herein), a UE 115 (e.g., any UE described herein), a network controller, a device, an equipment, a computing system, one or more components, or another suitable processing entity configured to perform any of the techniques described herein. For example, the node can be a UE 115. As another example, the node can be a network entity 105. As yet another example, a first node can be configured to communicate with a second node or a third node. In one aspect of this example, the first node can be a UE 115, the second node can be a network entity 105, and the third node can be a UE 115. In another aspect of this example, the first node can be a UE 115, the second node can be a network entity 105, and the third node can be a network entity 105. In other aspects of this example, the first node, the second node, and the third node can be different from these examples. Similarly, references to UEs 115, network entities 105, devices, equipment, computing systems, etc. may include the disclosure of UEs 115, network entities 105, devices, equipment, computing systems, etc. as nodes. For example, the disclosure that a UE 115 is configured to receive information from a network entity 105 also discloses that a first node is configured to receive information from a second node.

[0069] In some examples, network entity 105 may communicate with core network 130 or with each other or both. For example, network entity 105 may communicate with core network 130 via one or more backhaul communication links 120 (e.g., according to S1, N2, N3, or other interface protocols). In some examples, network entity 105 may communicate with each other directly (e.g., directly between network entities 105) or indirectly (e.g., via core network 130) via backhaul communication links 120 (e.g., according to X2, Xn, or other interface protocols). In some examples, network entity 105 may communicate with each other via midhaul communication link 162 (e.g., according to midhaul interface protocol) or fronthaul communication link 168 (e.g., according to fronthaul interface protocol) or any combination thereof. Backhaul communication link 120, midhaul communication link 162, or fronthaul communication link 168 may be or include one or more wired links (e.g., electrical link, optical fiber link), one or more wireless links (e.g., radio link, wireless optical link), etc. or various combinations thereof. UE 115 may communicate with core network 130 via communication link 155.

[0070] One or more of the network entities 105 described herein may include or may be referred to as base station 140 (e.g., transceiver base station, radio base station, NR base station, access point, radio transceiver, Node B, evolved Node B (eNB), next generation Node B, or gigabit Node B (any of which may be referred to as gNB), 5G NB, next generation eNB (ng-eNB), home Node B, home evolved Node B, or other suitable terms). In some examples, network entity 105 (e.g., base station 140) may be implemented in an aggregated (e.g., monolithic, stand-alone) base station architecture that may be configured to utilize a protocol stack physically or logically integrated within a single network entity 105 (e.g., a single RAN node, such as base station 140).

[0071] In some examples, network entity 105 may be implemented in a split architecture (e.g., split base station architecture, split RAN architecture), which may be configured to utilize a protocol stack physically or logically distributed between two or more network entities 105 (such as an integrated access backhaul (IAB) network, an open RAN (O-RAN) (e.g., a network configuration sponsored by the O-RAN Alliance), or a virtualized RAN (vRAN) (e.g., a cloud RAN (C-RAN))). For example, network entity 105 may include one or more of the following: a central unit (CU) 160, a distributed unit (DU) 165, a radio unit (RU) 170, a RAN intelligent controller (RIC) 175 (e.g., a near-real-time RIC (near RT RIC), a non-real-time RIC (non RT RIC)), a service management and orchestration (SMO) 180 system, or any combination thereof. The RU 170 may also be referred to as a radio headend, an intelligent radio headend, a remote radio headend (RRH), a remote radio unit (RRU), or a transmit receive point (TRP). One or more components of network entity 105 in the split RAN architecture may be co-located, or one or more components of network entity 105 may be located in distributed locations (e.g., separate physical locations). In some examples, one or more network entities 105 of the split RAN architecture may be implemented as virtual units (e.g., a virtual CU (VCU), a virtual DU (VDU), a virtual RU (VRU)).

[0072] The functional split between the CU 160, DU 165, and RU 170 is flexible and can support different functionalities, depending on which functions are performed at the CU 160, DU 165, or RU 170 (e.g., network layer functions, protocol layer functions, baseband functions, RF functions, and any combination thereof). For example, a functional split of the protocol stack can be adopted between the CU 160 and the DU 165 such that the CU 160 can support one or more layers of the protocol stack and the DU 165 can support one or more different layers of the protocol stack. In some examples, the CU 160 can host higher protocol layer (e.g., layer 3 (L3), layer 2 (L2)) functionality and signaling (e.g., radio resource control (RRC), service data adaptation protocol (SDAP), packet data convergence protocol (PDCP)). The CU 160 can be connected to one or more DU 165s or RU 170s, and one or more DU 165s or RU 170s can host lower protocol layers, such as layer 1 (L1) (e.g., physical (PHY) layer) or L2 (e.g., radio link control (RLC) layer, medium access control (MAC) layer) functionality and signaling, and can each be at least partially controlled by the CU 160. Additionally or alternatively, a functional split of the protocol stack can be adopted between the DU 165 and the RU 170 such that the DU 165 can support one or more layers of the protocol stack and the RU 170 can support one or more different layers of the protocol stack. The DU 165 can support one or more different cells (e.g., via one or more RU 170s). In some cases, the functional split between the CU 160 and the DU 165 or between the DU 165 and the RU 170 can be within a protocol layer (e.g., some functions of a protocol layer can be performed by one of the CU 160, DU 165, or RU 170, while other functions of that protocol layer are performed by a different one of the CU 160, DU 165, or RU 170). The CU 160 can be further functionally split into a CU control plane (CU-CP) and a CU user plane (CU-UP) function. The CU 160 can be connected to one or more DU 165s via an intermediate transport communication link 162 (e.g., F1, F1-c, F1-u), and the DU 165 can be connected to one or more RU 170s via a fronthaul communication link 168 (e.g., open fronthaul (FH) interface). In some examples, the intermediate transport communication link 162 or the fronthaul communication link 168 can be implemented according to the interfaces (e.g., channels) between the layers of the protocol stack, which are supported by the corresponding network entities 105 communicating via such communication links.

[0073] In some wireless communication systems (e.g., wireless communication system 100), the infrastructure and spectrum resources for radio access may support wireless backhaul link capabilities to supplement a wired backhaul connection and thereby provide an IAB network architecture (e.g., to a core network 130). In some cases, in an IAB network, one or more network entities 105 (e.g., IAB nodes 104) may be partially controlled by each other. One or more IAB nodes 104 may be referred to as donor entities or IAB donors. One or more DUs 165 or one or more RUs 170 may be partially controlled by one or more CUs 160 associated with a donor network entity 105 (e.g., donor base station 140). One or more donor network entities 105 (e.g., IAB donors) may communicate with one or more additional network entities 105 (e.g., IAB nodes 104) via the supported access and backhaul links (e.g., backhaul communication link 120). An IAB node 104 may include an IAB mobile terminal (IAB-MT) controlled (e.g., scheduled) by a DU 165 of a coupled IAB donor. The IAB-MT may include a separate antenna set for relaying communication with a UE 115 or may share the same antenna (e.g., of an RU 170 of the IAB node 104) for access via the DU 165 of the IAB node 104 (e.g., referred to as a virtual IAB-MT (vIAB-MT)). In some examples, an IAB node 104 may include a DU 165 that supports communication links with additional entities (e.g., IAB nodes 104, UEs 115) within a relay chain or configuration (e.g., downstream) of the access network. In such cases, one or more components of a split RAN architecture (e.g., one or more IAB nodes 104 or components of an IAB node 104) may be configured to operate in accordance with the techniques described herein.

[0074] For example, the access network (AN) or RAN may include communication between an access node (e.g., an IAB donor), an IAB node 104, and one or more UEs 115. The IAB donor may facilitate the connection between the core network 130 and the AN (e.g., via a wired or wireless connection to the core network 130). That is, the IAB donor may refer to a RAN node having a wired or wireless connection to the core network 130. The IAB donor may include a CU 160 and at least one DU 165 (e.g., and a RU 170), where the CU 160 may communicate with the core network 130 via an interface (e.g., a fronthaul link). The IAB donor and the IAB node 104 may communicate via an F1 interface according to a protocol that defines signaling messages (e.g., the F1 AP protocol). Additionally or alternatively, the CU 160 may communicate with the core network via an interface (which may be an example of a part of the fronthaul link), and may communicate with other CUs 160 (e.g., CUs 160 associated with alternative IAB donors) via an Xn-C interface (which may be an example of a part of the fronthaul link).

[0075] The IAB node 104 may refer to a RAN node that provides IAB functionality (e.g., access for UEs 115, wireless self-backhaul capabilities, etc.). The DU 165 may act as a distributed scheduling node towards the child nodes associated with the IAB node 104, and the IAB-MT may act as a scheduled node towards the parent node associated with the IAB node 104. That is, the IAB donor may be referred to as a parent node that communicates with one or more child nodes (e.g., the IAB donor may relay transmissions for UEs through one or more other IAB nodes 104). Additionally or alternatively, depending on the relay chain or configuration of the AN, the IAB node 104 may also be referred to as a parent node or a child node of other IAB nodes 104. Thus, the IAB-MT entity of the IAB node 104 may provide a Uu interface for a child IAB node 104 to receive signaling from a parent IAB node 104, and the DU interface (e.g., the DU 165) may provide a Uu interface for the parent IAB node 104 to signal to the child IAB node 104 or the UE 115.

[0076] For example, the IAB node 104 may be referred to as a parent node that supports communication for a sub-IAB node or as a sub-node associated with an IAB donor or both. The IAB donor may include a CU 160 having a wired or wireless connection (e.g., a fronthaul communication link 120) to the core network 130 and may act as the parent node of the IAB node 104. For example, the DU 165 of the IAB donor may relay transmissions to the UE 115 via the IAB node 104, or may signal transmissions directly to the UE 115, or both. The CU 160 of the IAB donor may signal communication link establishment to the IAB node 104 via the F1 interface, and the IAB node 104 may schedule transmissions (e.g., transmissions relayed from the IAB donor to the UE 115) via the DU 165. That is, data may be relayed to and from the IAB node 104 via signaling over the NR Uu interface to the MT of the IAB node 104. Communication with the IAB node 104 may be scheduled by the DU 165 of the IAB donor, and communication with the IAB node 104 may be scheduled by the DU 165 of the IAB node 104.

[0077] In the context where the techniques described herein are applied to a split RAN architecture, one or more components of the split RAN architecture may be configured to support model tuning for cross-node machine learning as described herein. For example, some operations described as being performed by the UE 115 or the network entity 105 (e.g., the base station 140) may additionally or alternatively be performed by one or more components of the split RAN architecture (e.g., the IAB node 104, the DU 165, the CU 160, the RU 170, the RIC 175, the SMO 180).

[0078] The UE 115 may include or may be referred to as a mobile device, a wireless device, a remote device, a handheld device, or a subscriber device or some other suitable term, where "device" may also be referred to as a unit, a station, a terminal, or a client, etc. The UE 115 may also include or may be referred to as a personal electronic device, such as a cellular phone, a personal digital assistant (PDA), a tablet computer, a laptop computer, or a personal computer. In some examples, the UE 115 may include or may be referred to as a wireless local loop (WLL) station, an Internet of Things (IoT) device, an Internet of Everything (IoE) device, or a machine type communication (MTC) device, etc., which may be implemented in various objects such as appliances or vehicles, meters, etc.

[0079] The UE 115 described herein may be capable of communicating with various types of devices such as other UEs 115 that may sometimes act as relays, as well as network entities 105 and network equipment including macro eNBs or gNBs, small cell eNBs or gNBs, or relay base stations, etc., as Figure 1as shown

[0080] UE 115 and network entity 105 may wirelessly communicate with each other via one or more communication links 125 (e.g., access links) using resources associated with one or more carriers. The term "carrier" may refer to a set of RF spectrum resources having a physical layer structure defined to support communication link 125. For example, a carrier for communication link 125 may include a portion (e.g., bandwidth part (BWP)) of an RF spectrum band operating according to one or more physical layer channels for a given radio access technology (e.g., LTE, LTE-A, LTE-A Pro, NR, advanced 5G). Each physical layer channel may carry acquisition signaling (e.g., synchronization signals, system information), control signaling for coordinating carrier operation, user data, or other signaling. Wireless communication system 100 may support communication with UE 115 using carrier aggregation or multi-carrier operation. According to a carrier aggregation configuration, UE 115 may be configured to have multiple downlink component carriers and one or more uplink component carriers. Carrier aggregation may be used for both frequency division duplex (FDD) and time division duplex (TDD) component carriers. Communication between network entity 105 and other devices may refer to communication between these devices and any part of network entity 105 (e.g., entity, sub-entity). For example, the terms "transmit", "receive", or "communicate" when referring to network entity 105 may refer to any part of network entity 105 of the RAN (e.g., base station 140, CU 160, DU 165, RU 170) communicating with another device (e.g., directly or via one or more other network entities 105).

[0081] In some examples, such as in a carrier aggregation configuration, a carrier may also have acquisition signaling or control signaling for coordinating the operation of other carriers. A carrier may be associated with a frequency channel (e.g., evolved universal mobile telecommunications system terrestrial radio access (E-UTRA) absolute RF channel number (EARFCN)) and may be identified according to a channel raster for discovery by UE 115. A carrier may operate in independent mode, in which case initial acquisition and connection may be performed by UE 115 via the carrier, or a carrier may operate in non-independent mode, in which case the connection is anchored using a different carrier (e.g., different carriers of the same or different radio access technologies).

[0082] The communication link 125 shown in the wireless communication system 100 may include a downlink transmission (e.g., forward link transmission) from the network entity 105 to the UE 115, an uplink transmission (e.g., reverse link transmission) from the UE 115 to the network entity 105, or other transmission configurations such as both. A carrier may carry downlink communication or uplink communication (e.g., in FDD mode), or may be configured to carry both downlink communication and uplink communication (e.g., in TDD mode).

[0083] A carrier may be associated with a particular bandwidth of the RF spectrum, and in some examples, the carrier bandwidth may be referred to as the "system bandwidth" of the carrier or the wireless communication system 100. For example, the carrier bandwidth may be one of a set of bandwidths of carriers of a particular radio access technology (e.g., 1.4 megahertz (MHz), 3 MHz, 5 MHz, 10 MHz, 15 MHz, 20 MHz, 40 MHz, or 80 MHz). Devices of the wireless communication system 100 (e.g., the network entity 105, the UE 115, or both) may have a hardware configuration that supports communication using a particular carrier bandwidth, or may be configurable to support communication using one of a set of carrier bandwidths. In some examples, the wireless communication system 100 may include a network entity 105 or a UE 115 that supports concurrent communication using carriers associated with multiple carrier bandwidths. In some examples, each served UE 115 may be configured to operate using a portion (e.g., a subband, a BWP) or all of the carrier bandwidth.

[0084] The signal waveform transmitted via a carrier may include multiple subcarriers (e.g., using a multicarrier modulation (MCM) technique such as orthogonal frequency division multiplexing (OFDM) or discrete Fourier transform spread OFDM (DFT-S-OFDM)). In a system employing an MCM technique, a resource element may refer to the resource of one symbol period (e.g., the duration of one modulation symbol) and one subcarrier, in which case the symbol period and the subcarrier spacing may be inversely related. The number of bits carried by each resource element may depend on the modulation scheme (e.g., the order of the modulation scheme, the decoding rate of the modulation scheme, or both), such that a relatively high number of resource elements (e.g., during the transmission duration) and a relatively high-order modulation scheme may correspond to a relatively high communication rate. Wireless communication resources may refer to a combination of RF spectrum resources, time resources, and spatial resources (e.g., spatial layers or beams), and the use of multiple spatial resources may increase the data rate or data integrity for communication with the UE 115.

[0085] One or more parameter sets may be supported for a carrier, and the parameter set may include the subcarrier spacing ( ( ) and a cyclic prefix. A carrier can be divided into one or more BWPs with the same or different parameter sets. In some examples, the UE 115 can be configured with multiple BWPs. In some examples, a single BWP of a carrier can be active at a given time, and the communication of the UE 115 can be restricted to one or more active BWPs.

[0086] A time interval for the network entity 105 or the UE 115 can be expressed as a multiple of a basic time unit, and the basic time unit can refer to, for example, a sampling period seconds, for which can represent the supported subcarrier spacing, and can represent the supported discrete Fourier transform (DFT) size. The time intervals of the communication resources can be organized according to radio frames each having a specified duration (e.g., 10 milliseconds (ms)). Each radio frame can be identified by a system frame number (SFN) (e.g., ranging from 0 to 1023).

[0087] Each frame can include a plurality of consecutively numbered subframes or time slots, and each subframe or time slot can have the same duration. In some examples, a frame can be divided (e.g., in the time domain) into subframes, and each subframe can be further divided into a certain number of time slots. Alternatively, each frame can include a variable number of time slots, and the number of time slots can depend on the subcarrier spacing. Each time slot can include a certain number of symbol periods (e.g., depending on the length of the cyclic prefix appended to each symbol period). In some wireless communication systems 100, a time slot can be further divided into a plurality of mini - slots associated with one or more symbols. Excluding the cyclic prefix, each symbol period can be associated with one or more (e.g., number of) sampling periods. The duration of a symbol period can depend on the subcarrier spacing or the operating frequency band.

[0088] A subframe, a time slot, a mini - slot, or a symbol can be the smallest scheduling unit of the wireless communication system 100 (e.g., in the time domain), and can be referred to as a transmission time interval (TTI). In some examples, the TTI duration (e.g., the number of symbol periods in a TTI) can be variable. Additionally or alternatively, the smallest scheduling unit of the wireless communication system 100 can be dynamically selected (e.g., in a burst of shortened TTIs (sTTIs)).

[0089] According to various techniques, carriers can be used to multiplex physical channels for communication. For example, one or more of time-division multiplexing (TDM) techniques, frequency-division multiplexing (FDM) techniques, or hybrid TDM-FDM techniques can be used to multiplex physical control channels and physical data channels for signaling via a downlink carrier. The control region of a physical control channel (e.g., a control resource set (CORESET)) can be defined by a set of symbol periods and can extend across the system bandwidth of a carrier or a subset of that system bandwidth. One or more control regions (e.g., CORESETs) can be configured for a set of UEs 115. For example, one or more of the UEs 115 can monitor or search a control region for control information according to one or more search space sets, and each search space set can include one or more control channel candidates in one or more aggregation levels arranged in a cascaded manner. The aggregation level of a control channel candidate can refer to the amount of control channel resources (e.g., control channel elements (CCEs)) associated with the coded information for a control information format with a given payload size. The search space sets can include: a common search space set configured to transmit control information to a plurality of UEs 115, and a UE-specific search space set for transmitting control information to a specific UE 115.

[0090] The network entity 105 can provide communication coverage via one or more cells (e.g., macro cells, small cells, hotspots, or other types of cells or any combination thereof). The term "cell" can refer to a logical communication entity for communicating with the network entity 105 (e.g., using a carrier) and can be associated with an identifier (e.g., a physical cell identifier (PCID), a virtual cell identifier (VCID), or other cell identifier) for distinguishing adjacent cells. In some examples, a cell can also refer to a coverage area 110 or a portion of the coverage area 110 (e.g., a sector) on which the logical communication entity operates. Depending on various factors such as the capabilities of the network entity 105, the scope of such cells can range from a smaller area (e.g., a structure, a subset of a structure) to a larger area. For example, a cell can be or can include a building, a subset of a building, or an external space between or overlapping the coverage areas 110, etc.

[0091] Macro cells generally cover a relatively large geographical area (e.g., with a radius of several kilometers) and may allow unrestricted access to UEs 115 that have a service subscription with the network provider that supports the macro cell. Compared with macro cells, small cells may be associated with lower-power network entities 105 (e.g., lower-power base stations 140), and small cells may operate using the same or different (e.g., licensed, unlicensed) frequency bands as macro cells. Small cells may provide unrestricted access to UEs 115 that have a service subscription with the network provider, or may provide restricted access to UEs 115 that are associated with the small cell (e.g., UEs 115 in a closed subscriber group (CSG), UEs 115 associated with users in a home or office). Network entity 105 may support one or more cells and may also use one or more component carriers to support communication via one or more cells.

[0092] In some examples, a carrier may support multiple cells and may be configured with different cells according to different protocol types that may provide access for different types of devices (e.g., MTC, narrowband IoT (NB-IoT), enhanced mobile broadband (eMBB)).

[0093] In some examples, network entity 105 (e.g., base station 140, RU 170) may be movable and thus provide communication coverage for a moving coverage area 110. In some examples, different coverage areas 110 associated with different technologies may overlap, but different coverage areas 110 may be supported by the same network entity 105. In some other examples, overlapping coverage areas 110 associated with different technologies may be supported by different network entities 105. The wireless communication system 100 may include, for example, a heterogeneous network in which different types of network entities 105 use the same or different radio access technologies to provide coverage for various coverage areas 110.

[0094] The wireless communication system 100 may support synchronous or asynchronous operation. For synchronous operation, network entity 105 (e.g., base station 140) may have similar frame timings, and transmissions from different network entities 105 may be approximately aligned in time. For asynchronous operation, network entity 105 may have different frame timings, and in some examples, transmissions from different network entities 105 may not be aligned in time. The techniques described herein may be used for synchronous operation or asynchronous operation.

[0095] Some UEs 115 (such as MTC or IoT devices) can be low-cost or low-complexity devices and can provide automated communication between machines (e.g., via machine-to-machine (M2M) communication). M2M communication or MTC can refer to data communication technologies that allow devices to communicate with each other or with network entity 105 (e.g., base station 140) without human intervention. In some examples, M2M communication or MTC can include communication from devices with integrated sensors or meters to measure or obtain information and relay such information to a central server or application that uses the information or presents the information to a person interacting with the application. Some UEs 115 can be designed to collect information or enable automated behavior of machines or other devices. Examples of applications for MTC devices include: smart metering, inventory monitoring, water level monitoring, equipment monitoring, healthcare monitoring, wildlife monitoring, weather and geographical event monitoring, queue management and tracking, remote security sensing, physical access control, and transaction-based commercial charging.

[0096] Some UEs 115 can be configured to operate in power consumption-reducing modes, such as half-duplex communication (e.g., a mode that supports one-way communication via transmission or reception but not concurrent transmission and reception). In some examples, half-duplex communication can be performed at a reduced peak rate. Other energy-saving techniques for UEs 115 include: entering a power-saving deep sleep mode when not participating in active communication, operating with limited bandwidth (e.g., according to narrowband communication), or a combination of these techniques. For example, some UEs 115 can be configured to operate using a narrowband protocol type associated with a defined portion or range (e.g., a set of subcarriers or resource blocks (RBs)) within a carrier, within a guard band of the carrier, or outside the carrier.

[0097] Wireless communication system 100 can be configured to support ultra-reliable communication or low-latency communication or various combinations thereof. For example, wireless communication system 100 can be configured to support ultra-reliable low-latency communication (URLLC). UEs 115 can be designed to support ultra-reliable, low-latency, or critical functions. Ultra-reliable communication can include private communication or group communication and can be supported by one or more services (such as push-to-talk, video, or data). Support for ultra-reliable, low-latency functions can include prioritization of services, and such services can be used for public safety or general commercial applications. The terms ultra-reliable, low-latency, and ultra-reliable low-latency can be used interchangeably herein.

[0098] In some examples, UE 115 may be configured to support communicating directly with other UEs 115 via a device-to-device (D2D) communication link 135 (e.g., according to a peer-to-peer (P2P), D2D, or sidelink protocol). In some examples, one or more UEs 115 performing D2D communication in a group may be within the coverage area 110 of a network entity 105 (e.g., base station 140, RU 170), and the network entity may support aspects of such D2D communication configured (e.g., scheduled) by the network entity 105. In some examples, one or more UEs 115 in such a group may be outside the coverage area 110 of the network entity 105, or may otherwise be unable or not configured to receive transmissions from the network entity 105. In some examples, a group of UEs 115 communicating via D2D communication may support a one-to-many (1:M) system, where each UE 115 transmits to each of the other UEs 115 in the group. In some examples, the network entity 105 may facilitate the scheduling of resources for D2D communication. In some other examples, D2D communication may be performed between UEs 115 without involving the network entity 105.

[0099] In some systems, the D2D communication link 135 may be an example of a communication channel (such as a sidelink communication channel) between vehicles (e.g., UE 115). In some examples, vehicles may communicate using vehicle-to-everything (V2X) communication, vehicle-to-vehicle (V2V) communication, or some combination of these. Vehicles may signal information related to traffic conditions, signal scheduling, weather, safety, emergencies, or any other information related to the V2X system. In some examples, vehicles in a V2X system may communicate with roadside infrastructure (such as a roadside unit), or communicate with the network via one or more network nodes (e.g., network entity 105, base station 140, RU 170) using vehicle-to-network (V2N) communication, or both.

[0100] The core network 130 can provide user authentication, access authorization, tracking, Internet Protocol (IP) connectivity, and other access, routing, or mobility functions. The core network 130 can be an evolved packet core (EPC) or a 5G core (5GC), which can include at least one control plane entity for managing access and mobility (e.g., a mobility management entity (MME), an access and mobility management function (AMF)) and at least one user plane entity for routing packets or interconnecting to an external network (e.g., a serving gateway (S-GW), a packet data network (PDN) gateway (P-GW), or a user plane function (UPF)). The control plane entity can manage non-access stratum (NAS) functions, such as the mobility, authentication, and bearer management of the UE 115 served by a network entity 105 (e.g., a base station 140) associated with the core network 130. User IP packets can be passed through the user plane entity, which can provide IP address allocation and other functions. The user plane entity can be connected to the IP services 150 of one or more network operators. The IP services 150 can include access to the Internet, an intranet, an IP multimedia subsystem (IMS), or packet switched streaming services.

[0101] The wireless communication system 100 can operate using one or more frequency bands in the range of 300 megahertz (MHz) to 300 gigahertz (GHz). Generally, the region from 300 MHz to 3 GHz is known as the ultra-high frequency (UHF) region or the decimeter band, because in terms of length, the wavelength range is from approximately one decimeter to one meter. UHF waves can be blocked or redirected by buildings and environmental features (which can be referred to as clusters), but these waves can be sufficient to penetrate structures so that macro cells can serve UEs 115 located indoors. Compared with communications using smaller frequencies and longer wavelengths in the high frequency (HF) or very high frequency (VHF) portions of the spectrum below 300 MHz, communications using UHF waves can be associated with smaller antennas and shorter ranges (e.g., less than 100 kilometers).

[0102] The wireless communication system 100 may also operate using the Super High Frequency (SHF) region (also known as the centimeter band) in the range of 3 GHz to 30 GHz or using the Extremely High Frequency (EHF) region of the spectrum (e.g., 30 GHz to 300 GHz) (also known as the millimeter band). In some examples, the wireless communication system 100 may support millimeter wave (mmW) communication between the UE 115 and the network entity 105 (e.g., the base station 140, the RU 170), and the EHF antennas of the corresponding devices may be smaller and closer spaced than UHF antennas. In some examples, such techniques may facilitate the use of antenna arrays within the device. However, the propagation of EHF transmissions may be affected by greater attenuation and shorter range than SHF or UHF transmissions. The techniques disclosed herein may be employed across transmissions using one or more different frequency regions, and the use of frequency bands designated across these frequency regions may vary by country or regulatory authority.

[0103] The wireless communication system 100 may utilize licensed and unlicensed RF spectrum bands. For example, the wireless communication system 100 may use an unlicensed band (such as the 5 GHz Industrial, Scientific, and Medical (ISM) band) to employ Licensed-Assisted Access (LAA), Long-Term Evolution Unlicensed (LTE-U) radio access technology, or NR technology. When operating using an unlicensed RF spectrum band, devices such as the network entity 105 and the UE 115 may employ carrier sensing for collision detection and avoidance. In some examples, the operation using the unlicensed band may be based on a carrier aggregation configuration (e.g., LAA) in combination with the operation using a licensed band component carrier. The operation using the unlicensed spectrum may include downlink transmissions, uplink transmissions, peer-to-peer (P2P) transmissions, device-to-device (D2D) transmissions, and so on.

[0104] The network entity 105 (e.g., the base station 140, the RU 170) or the UE 115 may be equipped with multiple antennas that may be used to employ techniques such as transmit diversity, receive diversity, multiple-input multiple-output (MIMO) communication, or beamforming. The antennas of the network entity 105 or the UE 115 may be located within one or more antenna arrays or antenna panels, which may support MIMO operation or transmit or receive beamforming. For example, one or more base station antennas or antenna arrays may be co-located at an antenna assembly (such as an antenna tower). In some examples, the antennas or antenna arrays associated with the network entity 105 may be located at different geographical locations. The network entity 105 may include an antenna array having a set of multiple rows and multiple columns of antenna ports that the network entity 105 may use to support beamforming for communication with the UE 115. Similarly, the UE 115 may include one or more antenna arrays that may support various MIMO or beamforming operations. Additionally or alternatively, the antenna panel may support RF beamforming for signals transmitted via the antenna ports.

[0105] The network entity 105 or the UE 115 may use MIMO communication to utilize multipath signal propagation and improve spectral efficiency by transmitting or receiving multiple signals via different spatial layers. Such techniques may be referred to as spatial multiplexing. The multiple signals may be transmitted, for example, by the transmitting device via different antennas or different combinations of antennas. Similarly, the multiple signals may be received by the receiving device via different antennas or different combinations of antennas. Each of the multiple signals may be referred to as a separate spatial stream and may carry information associated with the same data stream (e.g., the same codeword) or different data streams (e.g., different codewords). The different spatial layers may be associated with different antenna ports for channel measurement and reporting. MIMO techniques include: single-user MIMO (SU-MIMO), where multiple spatial layers are transmitted to the same receiving device; and multi-user MIMO (MU-MIMO), where multiple spatial layers are transmitted to multiple devices.

[0106] Beamforming (which may also be referred to as spatial filtering, directional transmission, or directional reception) is a signal processing technique that may be used at a transmitting device or a receiving device (e.g., the network entity 105, the UE 115) to shape or steer an antenna beam (e.g., a transmit beam, a receive beam) along a spatial path between the transmitting device and the receiving device. Beamforming may be achieved by combining signals conveyed via the antenna elements of an antenna array such that some signals propagating along a particular direction relative to the antenna array experience constructive interference while other signals experience destructive interference. The adjustment of the signals conveyed via the antenna elements may include the transmitting device or the receiving device applying an amplitude offset, a phase offset, or both to the signals carried via the antenna elements associated with the device. The adjustment associated with each of these antenna elements may be defined by a set of beamforming weights associated with a particular direction (e.g., relative to the antenna array of the transmitting device or the receiving device or relative to some other direction).

[0107] The network entity 105 or the UE 115 may use beam scanning techniques as part of a beamforming operation. For example, the network entity 105 (e.g., the base station 140, the RU 170) may use multiple antennas or antenna arrays (e.g., an antenna panel) to perform a beamforming operation for directional communication with the UE 115. Some signals (e.g., synchronization signals, reference signals, beam selection signals, or other control signals) may be transmitted by the network entity 105 multiple times in different directions. For example, the network entity 105 may transmit signals according to different sets of beamforming weights associated with different transmission directions. The transmission along different beam directions may be used to identify (e.g., by the transmitting device such as the network entity 105, or by the receiving device such as the UE 115)) the beam directions for later transmission or reception by the network entity 105.

[0108] Some signals, such as data signals associated with a particular receiving device, may be transmitted by a transmitting device (e.g., transmitting network entity 105, transmitting UE 115) along a single beam direction (e.g., a direction associated with a receiving device such as receiving network entity 105 or receiving UE 115). In some examples, the beam direction associated with the transmission along a single beam direction may be determined based on signals transmitted along one or more beam directions. For example, UE 115 may receive one or more of the signals transmitted by network entity 105 in different directions, and may report to network entity 105 an indication of the signal that UE 115 receives with the highest signal quality or other acceptable signal quality.

[0109] In some examples, transmissions performed by a device (e.g., by network entity 105 or UE 115) may be carried out using multiple beam directions, and the device may use a combination of digital precoding or beamforming to generate a combined beam for transmission (e.g., from network entity 105 to UE 115). UE 115 may report feedback indicating precoding weights for one or more beam directions, and the feedback may correspond to a configured set of beams across the system bandwidth or one or more sub-bands. Network entity 105 may transmit reference signals (e.g., cell-specific reference signal (CRS), channel state information reference signal (CSI-RS)), which may or may not be precoded. UE 115 may provide feedback for beam selection, which may be a precoding matrix indicator (PMI) or codebook-based feedback (e.g., multi-panel type codebook, linear combination type codebook, port selection type codebook). Although these techniques are described with reference to signals transmitted by network entity 105 (e.g., base station 140, RU 170) in one or more directions, UE 115 may use similar techniques for transmitting signals multiple times in different directions (e.g., for identifying beam directions used by UE 115 for subsequent transmissions or receptions), or for transmitting signals in a single direction (e.g., for transmitting data to a receiving device).

[0110] The receiving device (e.g., UE 115) may perform receiving operations according to multiple receiving configurations (e.g., directional listening) when receiving various signals (such as synchronization signals, reference signals, beam selection signals, or other control signals) from a receiving device (e.g., network entity 105). For example, the receiving device may perform receiving according to multiple receiving directions by: receiving via different antenna sub-arrays, processing the received signals according to different antenna sub-arrays, receiving according to different sets of receive beamforming weights (e.g., different directional listening weight sets) applied to the signals received at multiple antenna elements of the antenna array, or processing the received signals according to different sets of receive beamforming weights applied to the signals received at multiple antenna elements of the antenna array, any of which may be referred to as "listening" according to different receiving configurations or receiving directions. In some examples, the receiving device may use a single receiving configuration to receive along a single beam direction (e.g., when receiving a data signal). The single receiving configuration may be aligned along a beam direction determined based on listening according to different receiving configuration directions (e.g., a beam direction determined to have the highest signal strength, highest signal-to-noise ratio (SNR), or other acceptable signal quality based on listening according to multiple beam directions).

[0111] The wireless communication system 100 may be a packet-based network operating according to a layered protocol stack. In the user plane, the communication at the bearer or PDCP layer may be IP-based. The RLC layer may perform packet segmentation and reassembly for conveyance via logical channels. The MAC layer may perform priority handling and multiplexing from logical channels into transport channels. The MAC layer may also implement error detection techniques, error correction techniques, or both to support retransmission to improve link efficiency. In the control plane, the RRC layer may provide the establishment, configuration, and maintenance of an RRC connection for radio bearers supporting user plane data between the UE 115 and the network entity 105 or the core network 130. The PHY layer may map the transport channels to physical channels.

[0112] UE 115 and network entity 105 may support retransmission of data to increase the likelihood that the data is successfully received. Hybrid Automatic Repeat reQuest (HARQ) feedback is a technique for increasing the likelihood of correctly receiving data via a communication link (e.g., communication link 125, D2D communication link 135). HARQ may include a combination of error detection (e.g., using Cyclic Redundancy Check (CRC)), Forward Error Correction (FEC), and retransmission (e.g., Automatic Repeat reQuest (ARQ)). HARQ may improve throughput at the MAC layer under poor radio conditions (e.g., low signal-to-noise ratio conditions). In some examples, a device may support same-slot HARQ feedback, in which case the device may provide HARQ feedback for data received in a previous symbol in a particular slot during that slot. In some other examples, the device may provide HARQ feedback in a subsequent slot or according to some other time interval.

[0113] Wireless communication system 100 may implement a cross-node machine learning model tuning (e.g., fine-tuning) process for a first device (such as UE 115) to communicate with a second device (such as network entity 105). In some examples, UE 115 may perform a tuning process of a machine learning model based on data samples (e.g., a first data set). The machine learning model may be a first machine learning model (e.g., a baseline machine learning model) configured at UE 115 by network entity 105 (e.g., deployed at UE 115 or indicated to UE 115 via a message sent by network entity 105). The first data set may include a collection of data samples collected by UE 115, network entity 105, or both. The training data set for the first machine learning model may be associated with a first distribution of samples, and the collection of the collected data samples may be associated with a second distribution of samples. The difference between the distributions of the samples (e.g., the magnitude of the inconsistency between the distributions) may be used to improve the performance (such as interference performance) and accuracy of the first machine learning model.

[0114] For example, UE 115 may use the first data set to perform a tuning process of the first machine learning model, and the difference in the distributions included in the first data set may improve the machine learning model. UE 115 may use the first data set to perform encoder tuning, decoder tuning, or both, where the encoder is located at UE 115 and the decoder is located at network entity 105. For example, the tuning process may be defined by Equations 1 and 2:

[0115]

[0116] where corresponds to the machine learning model of the encoder parameterized by at UE 115, corresponds to the encoded channel state, corresponding to the channel state at the encoder, corresponding to the machine learning model of the decoder parameterized at network entity 105, and corresponding to the decoded channel state. The performance of the machine learning model can be defined by a loss function. For example, the loss function can be defined by Equation 3 and an example of the loss function can be defined by Equation 4: where

[0117]

[0118] corresponds to the overall performance of the machine learning model, corresponds to the fidelity performance of the machine learning model, and corresponds to the regularization performance of the machine learning model.

[0119] For example, UE 115 can use a machine learning model (which uses an encoder) to encode the channel state, and UE 115 can report the encoded channel state to network entity 105. Network entity 105 can reconstruct the decoded channel state based on the received encoded channel state. The quality of the decoded channel state can be associated with the tuning process at the UE. In some examples, network entity 105 can enable UE 115 to perform the tuning process using the encoder (e.g., network entity 105 can send signaling that allows UE 115 to perform the tuning process). In this example, network entity 105 can jointly tune the encoder and decoder (e.g., as an encoder-decoder pair) or tune the decoder separately using a first data set. Additionally or alternatively, network entity 105 can request UE 115 to tune the encoder using the first data set. However, signaling that enables UE 115 to tune the encoder may result in increased latency and reduced communication efficiency.

[0120] According to the techniques herein, UE 115 can autonomously perform a tuning process using a first data set to generate a second machine learning model for use at UE 115, network entity 105, or both. For example, UE 115 can be allowed to tune the encoder without permission or explicit signaling from the network entity. Thus, UE 115 can use the information it has about the decoder to tune the encoder. If UE 115 does not share the information about the autonomously tuned encoder with network entity 105, network entity 105 can use a stable encoder to tune the decoder. In some examples, UE 115 can additionally tune the decoder and convey this information to network entity 105, or alternatively, UE 115 may not have information about the decoder and may not be able to tune the decoder.

[0121] ​In some examples, the UE 115 sends a capabilities message to a network entity. In some examples, the capabilities message can be a UE capabilities signal or message that includes a list of supported encoders (e.g., if multiple encoders are available for a given task). Additionally or alternatively, the UE 115 can send machine learning model selection signaling to the network entity 105 that indicates the machine learning model selected or to be selected by the UE 115 (e.g., if multiple machine learning models are available at the UE 115). In some other examples, the capabilities message can indicate whether the UE 115 can autonomously perform a tuning process. The tuning process can be one of an online tuning process or an offline tuning process. The offline tuning process can involve the UE 115 training a separate instance of an encoder or decoder while leaving the encoder or decoder used for inference unaffected during training. In some examples, the separate instance of the encoder or decoder can be stored at the UE 115 or in a cloud storage device (e.g., a network-accessible server that stores data). The offline tuning can be UE-specific or specific to the manufacturer or vendor of the UE 115. The online tuning process can involve training an encoder or decoder for interference. The online tuning can be performed when the UE 115 requests activation of the online tuning process. However, in some cases, the UE 115 can reject an activation command from the network entity 105, or the network entity 105 can request that the UE 115 fallback to a baseline encoder. The UE 115 can generate a second training dataset based on the tuning process, and the UE 115 or the network entity 105 can use the second training dataset to perform a tuning process of a second machine learning model for use at the corresponding encoder, decoder, or both.

[0122] Figure 2 Illustrates an example of a wireless communication system 200 that supports model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure. In some examples, the wireless communication system 200 can implement aspects of the wireless communication system 100. For example, the wireless communication system 200 includes a network entity 105-a and a UE 115-a in a coverage area 110-a, which can be examples of the corresponding devices described with reference Figure 1 above.

[0123] In some examples, the network entity 105-a can enable the UE 115-a to perform a tuning process using an encoder (e.g., the network entity 105 can send signaling that allows the UE 115-a to perform the tuning process). Additionally or alternatively, the network entity 105-a can request that the UE 115-a tune the encoder using a first dataset. However, the signaling that enables the UE 115-a to tune the encoder can result in increased latency and reduced communication efficiency.

[0124] In some examples, UE 115-a may primarily use a first data set to perform a tuning process to generate a second machine learning model (e.g., an updated machine learning model) for use at UE 115-a, network entity 105-a, or both. For example, UE 115-a may tune an encoder, and network entity 105-a may use the encoded channel state of the trained encoder to tune a decoder. In some examples, UE 115-a may provide signaling to network entity 105-a to tune the decoder at network entity 105-a.

[0125] Wireless communication system 200 may support a process for performing a tuning process of a machine learning model. For example, UE 115-a may use uplink communication link 205 to send data samples 210 (e.g., a first data set) to network entity 105-a. Network entity 105-a may collect data samples 210 from a set of UEs for performing a tuning process of a machine learning model. Network entity 105-a may use downlink communication link 215 to send a baseline update 220 to UE 115-a, and the baseline update may be sent via broadcast signaling or dedicated signaling for UE 115-a. Baseline update 220 may be an update to a first machine learning model (e.g., a baseline machine learning model) at UE 115-a. UE 115-a may use baseline update 220 to tune the first machine learning model and use the updated machine learning model (e.g., a second machine learning model) to perform communication with network entity 105-a.

[0126] In some examples, UE 115-a may prohibit performing a tuning process for an encoder and a decoder, and network entity 105-a may perform a tuning process for the encoder and the decoder. For example, network entity 105-a may use information from multiple UEs 115 to collect data samples 210, and may not utilize signaling (e.g., backpropagation signaling) for UE 115-a to perform a tuning process. Thus, network entity 105-a may use data samples 210 to update the encoder, and network entity 105-a may send channel state information (e.g., baseline update 220) to UE 115-a. Network entity 105-a may perform a tuning process in a timely manner (e.g., when the network is not fully utilized).

[0127] In this example, network entity 105-a may use one or more encoder and decoder pairs, which may be custom pairs not defined in the set of operational procedures defined by an operational standard. Network entity 105-a may also collect data samples 210 (e.g., channel state information (CSI) data samples) for training or tuning the one or more encoder and decoder pairs. Network entity 105-a may send a baseline update 220 to UE 115-a via broadcast signaling or dedicated signaling, and the baseline update may indicate the updated encoder and decoder pairs. Baseline update 220 may include encoder coefficients, decoder coefficients, or both. In another example, baseline update 220 may include an indicator (e.g., a uniform resource locator (URL)) indicating encoder coefficients, decoder coefficients, or both.

[0128] In some examples, UE 115-a may perform an online tuning process, and network entity 105-a may prohibit the execution of the tuning process, as Figure 3 further described. In some examples, UE 115-a may autonomously perform an online tuning process, and network entity 105-a may prohibit the execution of the tuning process, as Figure 4 further described. In some examples, UE 115-a may perform an offline tuning process, and network entity 105-a may prohibit the execution of the tuning process, as Figure 5 further described. In some examples, UE 115-a may autonomously perform an offline tuning process, and network entity 105-a may prohibit the execution of the tuning process, as Figure 6 further described. In some examples, UE 115-a may perform a tuning process, and UE 115-a may not have information about the decoder of network entity 105-a, as Figure 7 further described. In some examples, UE 115-a and network entity 105-a may jointly perform an online tuning process for both the encoder and the decoder, as Figure 8 further described.

[0129] In some examples, a baseline machine learning model may be used by UE 115-a. The baseline model may be an example of a model defined in a communication standard indicated by network entity 105-a (e.g., downloaded by UE 115-a or sent from network entity 105-a to UE 115-a), or may be deployed by the vendor or manufacturer of UE 115-a. The baseline model may be fixed or updated over time.

[0130] The machine learning model can be tuned or fine-tuned by leveraging a training data set, which can be collected by UE 115-a or network entity 105-a. The tuning can improve the inference performance based on the magnitude of the inconsistency between the distribution of the training data set and the distribution of data samples observed by UE 115-a or network entity 105-a. The tuning process can involve forward propagation, in which a first device (e.g., UE 115-a) sends information to a second device (e.g., network entity 105-a) that the second device uses to tune a machine learning model at the second device, which can be used for communication between the first device and the second device. The tuning process can be a fine-tuning process that uses both forward propagation and backpropagation. Backpropagation can involve sending the information back from the second device to the first device after the first device provides some initial information to the second device. For example, UE 115-a can provide information associated with forward propagation to network entity 105-a, which can be a (v, z) pair, where v represents the channel state information at UE 115-a (e.g., CSI encoder or channel state feedback (CSF) encoder). In some cases, UE 115-a can provide information about the channel between UE 115-a and network entity 105-a, which is represented by z. For example, UE 115-a can encode information about the channel state v into z using an encoder at UE 115-a.

[0131] After UE 115-a sends information associated with forward propagation, network entity 105-a can send information associated with backpropagation to UE 115-a. For example, network entity 105-a can reconstruct the channel state from using a decoder at network entity 105-a. Network entity 105-a can send information about the gradient of the channel (such as where L represents the loss function). UE 115-a can update the encoder parameters (θ) based on the gradient. That is, UE 115-a can update one or more parameters of a baseline parameter set ( to fine-tune a fine-tuned model that utilizes a set of fine-tuning parameters ( ). Thus, the encoder can be updated from to to . These fine-tuning steps can be repeated until a stopping criterion (e.g., convergence, fixed number of steps, signaling to stop) occurs.

[0132] Fine-tuning can be performed in a timely manner. For example, to reduce the signaling overhead for exchanging forward propagation and backward propagation, fine-tuning can be performed when the network is not fully utilized (e.g., when the network is not fully utilized, an allowed signal can be sent from network entity 105-a to UE 115-a to trigger UE 115-a to perform fine-tuning. The fine-tuning can be encoder-specific, decoder-specific, or both).

[0133] Figure 3 Illustrates an example of a process flow 300 that supports model tuning for cross-node machine learning according to one or more aspects of the present disclosure. In some examples, process flow 300 may implement aspects of wireless communication system 100 and wireless communication system 200. Process flow 300 may include UE 115-b and network entity 105-b, which may be examples of UE 115 and network entity 105 as described herein with reference to Figure 1 and Figure 2 described UE 115 and network entity 105.

[0134] Process flow 300 may illustrate an example of a technique that enables UE 115-b to perform a tuning process of a machine learning model. For example, UE 115-b may perform an online tuning process, and network entity 105-b may prohibit the execution of the tuning process. UE 115-b and network entity 105-b may use radio resource control (RRC) signaling, medium access control control element (MAC-CE) signaling, or physical layer signaling (e.g., downlink control indication (DCI)) to convey Figure 3 any of the messages or signals described.

[0135] At 305, UE 115-b may send a capability message to network entity 105-b. The capability message may indicate whether UE 115-b has the ability to tune the encoder of UE 115-b (e.g., optimize or modify its parameters) (e.g., by tuning the encoder using CSI samples collected by UE 115-b). For example, UE 115-b may be associated with a certain distribution of wireless channel data samples (e.g., experience that certain distribution), and UE 115-b may use the UE-specific CSI data sample distribution to tune the encoder. The capability message may indicate whether UE 115-b has the ability to perform offline tuning, online tuning, or both.

[0136] At 310, in some examples, UE 115-b may send an activation request to network entity 105-b. The activation request may indicate a request to perform an online tuning process of the encoder of UE 115-b on UE 115-b, and the activation request may indicate that UE 115-b has an encoder available for the tuning process.

[0137] At 315, network entity 105-b may send an activation status to UE 115-b. In some examples, network entity 105-b may send the activation status in response to an activation request. The activation status may indicate or direct UE 115-b to perform an online tuning process. UE 115-b may accept or reject the activation status (e.g., by sending a response to the activation status). In some examples, the activation status may request UE 115-b to use a baseline machine learning model for the encoder.

[0138] At 320, UE 115-b may begin to perform an online tuning process based on the activation status. If UE 115-b accepts the activation of the online tuning process, UE 115-b may perform the online tuning process. The online tuning process may be associated with UE 115-b tuning an encoder by (e.g., on the fly) using data samples to update encoding parameters, where the data samples are based on data samples associated with a decoder at network entity 105-a. In such a case, UE 115-b may perform the online tuning process without using a separate instance of the decoder or encoder.

[0139] At 325, in some examples, UE 115-b may send a deactivation status to network entity 105-b. The deactivation status may indicate that UE 115-b may or will stop performing the online tuning process.

[0140] At 330, in some examples, UE 115-b may stop performing the online tuning process based on the sent deactivation status. The deactivation may be initiated by UE 115-b or network entity 105-b.

[0141] At 335, network entity 105-b may send a deactivation status to UE 115-b. The deactivation status may indicate to UE 115-b to stop performing the online tuning process.

[0142] At 340, UE 115-b may stop performing the online tuning process based on the received deactivation status.

[0143] Figure 4 Illustrates an example of process flow 400 that supports model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure. In some examples, process flow 400 may implement aspects of wireless communication system 100 and wireless communication system 200. Process flow 400 may include UE 115-c and network entity 105-c, which may be examples of UE 115 and network entity 105 as described herein with reference to Figure 1 and Figure 2 described.

[0144] The process flow 400 can illustrate an example of a technique that enables the UE 115-c to perform a tuning process of a machine learning model. For example, the UE 115-c can perform an online autonomous tuning process, and the network entity 105-b can prohibit the execution of the tuning process. The UE 115-c and the network entity 105-c can use one or more of RRC signaling, MAC-CE signaling, or physical layer signaling (e.g., DCI) to convey Figure 4 any of the signals or messages described.

[0145] At 405, the UE 115-c can send a capability message to the network entity 105-c. The capability message can indicate whether the UE 115-c has the ability to tune the encoder of the UE 115-c (e.g., optimize or modify its parameters) (e.g., by tuning the encoder using CSI samples collected by the UE 115-c). The capability message can indicate whether the UE 115-c has the ability to perform offline tuning, online tuning, or both.

[0146] At 410, in some examples, the network entity 105-c can send an allowed status to the network entity 105-c. The allowed status can indicate to the UE 115-c that the UE 115-c is allowed to perform an autonomous online tuning process.

[0147] At 415, the UE 115-c can send an activation status or a deactivation status to the network entity 105-c. The activation status can indicate that the UE 115-c can start the online autonomous tuning of the encoder by determining updated parameters for the encoder (e.g., based on CSI data samples). The deactivation status can indicate that the UE 115-c can stop the online autonomous tuning of the encoder. The activation status or the deactivation status can cause synchronization of the encoder (e.g., the state of the encoder) with the network entity 105-c.

[0148] At 420, UE 115-c may start to perform an online autonomous tuning process based on an active state or an inactive state. If UE 115-c determines the purpose of the online autonomous tuning process, UE 115-c may execute the process. The online tuning process may be associated with UE 115-c tuning an encoder using data samples obtained from network entity 105-c. This may be obtained based on reference signals received from network entity 105-c during an active communication session (e.g., while in operation). In this case, UE 115-c may not use a separate instance of a decoder or an encoder to perform the tuning process. In some examples, UE 115-c performing an autonomous online tuning process may interfere with network entity 105-c. For example, as a result of UE 115-c and network entity 105-c tuning the encoder and decoder, respectively, the overall performance of the tuning processes at UE 115-c and network entity 105-c may degrade. Accordingly, network entity 105-c may send signaling to UE 115-c to control the tuning process.

[0149] At 425, network entity 105-c sends an inadmissible state to UE 115-c. The inadmissible state may indicate that UE 115-c stops the autonomous tuning process.

[0150] At 430, UE 115-c may stop performing the online tuning process based on the received inactive state.

[0151] Figure 5 An example of a process flow 500 that supports model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure is illustrated. In some examples, process flow 500 may implement aspects of wireless communication system 100 and wireless communication system 200. Process flow 500 may include UE 115-d and network entity 105-d, which may be examples of UE 115 and network entity 105 as described herein with reference to Figure 1 and Figure 2 described.

[0152] Process flow 500 may illustrate an example of a technique that enables UE 115-d to perform a tuning process of a machine learning model. For example, UE 115-d may perform an offline tuning process, and network entity 105-d may prohibit the execution of the tuning process. UE115-d and network entity 105-d may use one or more of RRC signaling, MAC-CE signaling, or physical layer signaling (e.g., DCI) to convey Figure 5 any of the signals or messages described.

[0153] At 505, UE 115-d may send a capabilities message to network entity 105-d. The capabilities message may indicate whether UE 115-d has the ability to tune an encoder. For example, UE 115-d may be associated with a certain distribution of wireless channel data samples (e.g., experiencing that certain distribution), and UE 115-d may use the UE-specific CSI data sample distribution to tune the encoder. The capabilities message may indicate that UE 115-d has the ability to perform offline tuning, online tuning, or both.

[0154] At 510, UE 115-d may begin to perform an offline tuning process. If UE 115-d determines that an offline tuning process is needed, UE 115-d may perform the process. The offline tuning process may be associated with a separate instance of UE 115-d tuning a first encoder or a first encoder and decoder pair, and UE 115-d may prohibit tuning a second encoder or a second encoder and decoder pair at UE 115-d.

[0155] At 515, UE 115-d may send an indication to network entity 105-d. The indication may indicate that UE 115-d has an available encoder (e.g., a fine-tuned encoder) from the offline tuning process.

[0156] At 520, network entity 105-d may send an activation status to UE 115-d. In some examples, network entity 105-d may send the activation status in response to an activation request. The activation status may indicate that UE 115-d is permitted to use the fine-tuned encoder. UE 115-d may explicitly accept or reject the activation status. In some examples, the activation status may request that UE 115-d use a baseline machine learning model for the encoder.

[0157] At 525, UE 115-d may begin to use the fine-tuned encoder based on the activation status.

[0158] At 530, in some examples, UE 115-d may send a deactivation status to network entity 105-d. The deactivation status may indicate that UE 115-d is stopping use of the fine-tuned encoder and beginning to use a baseline machine learning model for the encoder.

[0159] At 535, in some examples, UE 115-d may stop using the offline fine-tuned encoder based on the sent deactivation status. The deactivation may be initiated by UE 115-d or network entity 105-d.

[0160] At 540, network entity 105-d may send a deactivation status to UE 115-d. The deactivation status may instruct UE 115-d to stop using the offline fine-tuned encoder.

[0161] At 545, UE 115-d may stop using the offline fine-tuning encoder based on the received deactivation status.

[0162] Figure 6 An example of a process flow 600 that supports model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure is illustrated. In some examples, process flow 600 may implement aspects of wireless communication system 100 and wireless communication system 200. Process flow 600 may include UE 115-e and network entity 105-e, which may be examples of UE 115 and network entity 105 as described herein with reference to Figure 1 and Figure 2 the described UE 115 and network entity 105.

[0163] Process flow 600 may illustrate an example of a technique that enables UE 115-e to perform a tuning process of a machine learning model. For example, UE 115-e performs an offline autonomous tuning process, and network entity 105-e prohibits performing the tuning process. UE 115-e and network entity 105-e may use one or more of RRC signaling, MAC-CE signaling, or physical layer signaling (e.g., DCI) to convey Figure 6 any of the signals or messages described.

[0164] At 605, UE 115-e may send a capability message to network entity 105-e. The capability message may indicate whether UE 115-e has the ability to tune (e.g., optimize) an encoder (e.g., by tuning the encoder using CSI samples collected by UE 115-e). The capability message may indicate that UE 115-e has the ability to perform offline tuning, online tuning, or both.

[0165] At 610, UE 115-e may start performing an offline autonomous tuning process. If UE 115-e determines that an offline autonomous tuning process is needed, UE 115-d performs the process. The offline autonomous tuning process may be associated with a separate instance of UE 115-e tuning a first encoder or a first encoder and decoder pair, and UE 115-e may prohibit tuning a second encoder or a second encoder and decoder pair at UE 115-e.

[0166] At 615, in some examples, UE 115-e may send an indication to network entity 105-e. The indication may indicate that UE 115-e has an available encoder (i.e., a fine-tuned encoder) from the offline tuning process.

[0167] At 620, in some examples, network entity 105-e may send an allow status to network entity 105-e. The allow status may indicate to UE 115-e that UE 115-e is allowed to use the fine-tuned encoder.

[0168] At 625, in some examples, the UE 115-e may send an active state or an inactive state to the network entity 105-e. The active state may indicate that the UE 115-e may start using the fine-tuning encoder. The inactive state may indicate that the UE 115-e may stop using the fine-tuning encoder. The active state or the inactive state may cause synchronization of the encoder (e.g., the state of the encoder) with the network entity 105-e.

[0169] At 630, the UE 115-c may start using the offline fine-tuning encoder. This may be based on the active state or the inactive state. In some examples, the UE 115-e performing an offline autonomous tuning process may interfere with the network entity 105-e. For example, as a result of the UE 115-e and the network entity 105-e tuning the encoder and the decoder respectively, the overall performance of the tuning processes at the UE 115-e and the network entity 105-e may degrade. Accordingly, the network entity 105-e may send signaling to the UE 115-c to control the tuning process.

[0170] At 635, the network entity 105-e sends a not-allowed state to the UE 115-e. The not-allowed state may indicate that the UE 115-e stops using the fine-tuning encoder.

[0171] At 640, the UE 115-e may stop using the fine-tuning encoder based on the received inactive state.

[0172] Figure 7 An example of a process flow 700 that supports model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure is illustrated. In some examples, the process flow 700 may implement aspects of the wireless communication system 100 and the wireless communication system 200. The process flow 700 may include a UE 115-f and a network entity 105-f, which may be examples of the UE 115 and the network entity 105 as described herein with reference to Figure 1 and Figure 2 The process flow 700 may illustrate an example of a technique that enables the UE 115-f to perform a tuning process of a machine learning model. For example, the UE 115-f may perform the tuning process, and the UE 115-f may not have information about the decoder of the network entity 105-f. The UE 115-f may have information about the encoder, and the information may be communicated by the network entity 105-f or configured by the network entity 105-f.

[0173] In some examples, since UE 115-f may not have information about the decoder, the tuning process may include forward propagation and backpropagation. Forward propagation may include a method for moving an input layer (such as a first data set) to an output layer (such as a decoded channel state). Backpropagation may include a method for moving the output layer to the input layer (e.g., moving the decoded channel state to the first data set). In this example, UE 115-f may provide information associated with forward propagation to network entity 105-f. For example, UE 115-f may send parameters and . Network entity 105-f may use the information associated with backpropagation to respond to UE 115-f. Network entity 105-f may send the gradient of the loss function (∇zL, where L corresponds to the loss function defined by Equation 3) derived from the decoded channel state. In some examples, UE 115-f may tune (e.g., update) the encoder based on this gradient. UE 115-f may tune the encoder θ until a stopping criterion occurs. The stopping criterion may include convergence of the gradient of the loss function, a limit on the number of tuning steps, or signaling for UE 115-f to stop the tuning process. In some examples, UE 115-f may perform the tuning process opportunistically. For example, to reduce signaling overhead from forward propagation and backpropagation, UE 115-f may perform the tuning process (e.g., UE 115-f may receive an allowed state) when network entity 105-f is not fully utilized, which will be described in more detail below.

[0174] At 705, in some examples, network entity 105-f may send loss function information (e.g., α) for UE 115-f for backpropagation. UE 115-f may use the loss function information as a one-time input for the gradient of the loss function.

[0175] At 710, UE 115-f may send parameters and for forward propagation.

[0176] At 715, network entity 105-f may use these parameters to perform forward propagation and derive , i.e., the decoded channel state at the decoder. Network entity 105-f may use the received parameters and and the derived parameter to perform the calculation of the loss function L.

[0177] At 720, network entity 105-f may use the loss function L and the parameter To perform gradient backpropagation of the decoder.

[0178] At 725, network entity 105-f may send parameters and L for backpropagation at UE 115-f.

[0179] At 730, UE 115-f may use these parameters to perform gradient backpropagation of the encoder.

[0180] At 735, based on the gradient backpropagation, UE 115-f may update (e.g., tune) the encoder θ until a stopping criterion is met. For example, the tuning of the encoder may be defined by Equation 5:

[0181]

[0182] where the partial derivative may be estimated at , corresponding to the number of tuning steps, corresponding to the partial derivative of the loss function, and corresponding to the step size. For a given , the number of tuning steps may be equal to or greater than one.

[0183] Figure 8 Illustrates an example of a process flow 800 that supports model tuning for cross-node machine learning according to one or more aspects of the present disclosure. The process flow 800 may include UE 115-g and network entity 105-g, which may be examples of UE 115 and network entity 105 as described herein with reference to Figure 1 and Figure 2 . The process flow 800 may illustrate an example of a technique that enables UE 115-g to perform a tuning process of a machine learning model. For example, UE 115-g and network entity 105-g may jointly perform an online tuning process for both the encoder and the decoder.

[0184] Jointly performing the online tuning process may include forward propagation and backpropagation. In this example, UE 115-g may provide information associated with forward propagation to network entity 105-g. For example, UE 115-g may send parameters and to network entity 105-g. Network entity 105-g may use these parameters to calculate the gradient (∇zL) of the loss function derived from the decoded channel state, and network entity 105-g may update the decoder parameter φ based on this gradient. In some examples, network entity 105-g may send information associated with backpropagation to UE 115-g.

[0185] For example, network entity 105-g may send parameters associated with the gradient of the loss function. In some examples, UE 115-g may compute the gradient and tune the encoder based on the gradient. UE 115-g may tune the encoder θ until a stopping criterion occurs. The stopping criterion may include convergence of the gradient of the loss function, a limit on the number of tuning steps, or signaling for UE 115-g to stop the tuning process. In this example, network entity 105-g may not have information related to the encoder at UE 115-g, but the encoder θ and decoder φ may be jointly tuned. In some examples, UE 115-g may perform the tuning process opportunistically. For example, to reduce signaling overhead from forward and backward propagation, UE 115-g may perform the tuning process when network entity 105-g is not fully utilized (e.g., UE 115-g may receive an allowed state), which will be described in more detail below.

[0186] At 805, network entity 105-f may send loss function information (e.g., α) of UE 115-g to UE 115-g for backpropagation. UE 115-g may use the loss function information as a one-time input for the gradient of the loss function.

[0187] At 810, UE 115-g may send parameters and to network entity 105-g for forward propagation.

[0188] At 815, network entity 105-g may use these parameters to perform forward propagation and derive , i.e., the decoded channel state at the decoder. Network entity 105-g may use the received parameters and as well as the derived parameter to perform the calculation of the loss function L.

[0189] At 820, network entity 105-f may use the loss function L and the parameter to perform gradient backpropagation of the decoder.

[0190] At 825, based on the gradient backpropagation, network entity 105-g may update (e.g., tune) the decoder φ until a stopping criterion occurs. For example, the tuning of the decoder may be defined by Equation 6:

[0191]

[0192] where the partial derivative may be estimated at , corresponds to the number of tuning steps, corresponds to the partial derivative of the loss function, and corresponds to the step size. For a given , the number of tuning steps can be equal to or greater than one.

[0193] At 830, network entity 105-g can send parameters and L for backpropagation at UE 115-g.

[0194] At 835, UE 115-g can use these parameters to perform gradient backpropagation of the encoder.

[0195] At 840, based on the gradient backpropagation, UE 115-f can update (e.g., tune) the encoder θ until a stop criterion occurs, as defined by Equation 5.

[0196] Figure 9 FIG. 900 illustrates a block diagram 900 of a device 905 that supports model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure. Device 905 may be an example of aspects of UE 115 as described herein. Device 905 may include a receiver 910, a transmitter 915, and a communication manager 920. Device 905 may also include a processor. Each of these components may communicate with each other (e.g., via one or more buses).

[0197] Receiver 910 may provide components for receiving information (such as packets, user data, control information, or any combination thereof associated with various information channels (e.g., control channels, data channels, information channels related to techniques for model tuning for cross-node machine learning)). The information may be delivered to other components of device 905. Receiver 910 may utilize a single antenna or an array of multiple antennas.

[0198] Transmitter 915 may provide components for transmitting signals generated by other components of device 905. For example, transmitter 915 may transmit information such as packets, user data, control information, or any combination thereof associated with various information channels (e.g., control channels, data channels, information channels related to model tuning). In some examples, transmitter 915 may be co-located with receiver 910 in a transceiver module. Transmitter 915 may utilize a single antenna or an array of multiple antennas.

[0199] Communication manager 920, receiver 910, transmitter 915, or various combinations or various components thereof may be examples of components for performing various aspects of model tuning for cross-node machine learning as described herein. For example, communication manager 920, receiver 910, transmitter 915, or various combinations or components thereof may support methods for performing one or more of the functions described herein.

[0200] In some examples, the communication manager 920, the receiver 910, the transmitter 915, or various combinations or components thereof may be implemented in hardware (e.g., in a communication management circuit). The hardware may include a processor, a digital signal processor (DSP), a central processing unit (CPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic device, a microcontroller, discrete gate or transistor logic, discrete hardware components, or any combination thereof configured to or otherwise supporting components for performing the functions described in this disclosure. In some examples, the processor and the memory coupled to the processor may be configured to perform one or more of the functions described herein (e.g., by the processor executing instructions stored in the memory).

[0201] Additionally or alternatively, in some examples, the communication manager 920, the receiver 910, the transmitter 915, or various combinations or components thereof may be implemented in code executed by a processor (e.g., implemented as communication management software or firmware). If implemented in code executed by a processor, the functions of the communication manager 920, the receiver 910, the transmitter 915, or various combinations or components thereof may be performed by a general purpose processor, a DSP, a CPU, an ASIC, an FPGA, a microcontroller, or any combination of these or other programmable logic devices configured to or otherwise supporting components for performing the functions described in this disclosure.

[0202] In some examples, the communication manager 920 may be configured to use or otherwise cooperate with the receiver 910, the transmitter 915, or both to perform various operations (e.g., receive, obtain, monitor, output, transmit). For example, the communication manager 920 may receive information from the receiver 910, convey information to the transmitter 915, or integrate in combination with the receiver 910, the transmitter 915, or both to obtain information, output information, or perform various other operations as described herein.

[0203] According to examples disclosed herein, communication manager 920 may support wireless communication at a UE. For example, communication manager 920 may be configured to or otherwise support components for obtaining data samples for a first machine learning model associated with a task at the UE, where a first set of parameters is associated with the first machine learning model. Communication manager 920 may be configured to or otherwise support components for sending a capability message indicating the UE's ability to perform a tuning process of the first machine learning model. Communication manager 920 may be configured to or otherwise support components for performing a tuning process of the first machine learning model based on the UE's ability to perform the tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model. Communication manager 920 may be configured to or otherwise support components for sending a message indicating at least a portion of the second set of parameters to a network entity based on performing the tuning process of the first machine learning model.

[0204] Additionally or alternatively, according to examples disclosed herein, communication manager 920 may support wireless communication at a UE using a first set of parameters associated with a first machine learning model for a task. For example, communication manager 920 may be configured to or otherwise support components for sending a capability message indicating the UE's ability to perform a tuning process of the first machine learning model, where the first set of parameters is associated with the first machine learning model. Communication manager 920 may be configured to or otherwise support components for performing a tuning process of the first machine learning model based on the UE's ability to perform the tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model. Communication manager 920 may be configured to or otherwise support components for sending a message indicating the second set of parameters to a network entity based on performing the tuning process of the first machine learning model. Communication manager 920 may be configured to or otherwise support components for receiving an allowed status indication from a network entity associated with the second set of parameters. Communication manager 920 may be configured to or otherwise support components for performing the task using the first set of parameters associated with the first machine learning model or the second set of parameters associated with the first machine learning model based on the received allowed status indication. Communication manager 920 may be configured to or otherwise support components for receiving a not-allowed status indication from a network entity associated with the second set of parameters. Communication manager 920 may be configured to or otherwise support components for performing the task using the first set of parameters associated with the first machine learning model based on the received not-allowed status indication.

[0205] Additionally or alternatively, according to examples as disclosed herein, the communication manager 920 may utilize a first set of parameters associated with a first machine learning model for a task to support wireless communication at a UE. For example, the communication manager 920 may be configured as or otherwise support a component for sending a capability message to a network entity indicating the UE's ability to perform an online tuning process of the first machine learning model, where the first set of parameters is associated with the first machine learning model. The communication manager 920 may be configured as or otherwise support a component for receiving a first indication from a network entity associated with the online tuning process, where the first indication includes an activation state or an allowance state. The communication manager 920 may be configured as or otherwise support a component for performing an online tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model based on the UE's ability to perform a tuning process of the first machine learning model and the received allowance state.

[0206] By including or configuring the communication manager 920 according to examples as described herein, a device 905 (e.g., a control receiver 910, a transmitter 915, the communication manager 920, or a combination thereof or a processor otherwise coupled thereto) may support techniques for reducing power consumption and more efficiently utilizing communication resources. For example, by sending a capability message to a network entity, a UE may notify the network entity of its ability to autonomously tune a second machine learning model. Autonomously performing the tuning process may cause the processor of the device 905 to more efficiently tune the second machine learning model and reduce latency in communications using the second machine learning model.

[0207] Figure 10 Block diagram 1000 illustrates a device 1005 supporting model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure. The device 1005 may be an example of aspects of the device 905 or the UE 115 as described herein. The device 1005 may include a receiver 1010, a transmitter 1015, and a communication manager 1020. The device 1005 may also include a processor. Each of these components may communicate with each other (e.g., via one or more buses).

[0208] The receiver 1010 may provide a component for receiving information (such as packets, user data, control information, or any combination thereof associated with various information channels (e.g., control channels, data channels, information channels related to techniques for cross-node machine learning model tuning)). The information may be delivered to other components of the device 1005. The receiver 1010 may utilize a single antenna or an array of multiple antennas.

[0209] Transmitter 1015 may provide components for transmitting signals generated by other components of device 1005. For example, transmitter 1015 may transmit information such as packets, user data, control information, or any combination thereof associated with various information channels (e.g., control channels, data channels, information channels related to model tuning for cross-node machine learning). In some examples, transmitter 1015 may be co-located with receiver 1010 in a transceiver module. Transmitter 1015 may utilize a single antenna or an array of multiple antennas.

[0210] Device 1005 or its various components may be examples of components for performing various aspects of model tuning for cross-node machine learning as described herein. For example, communication manager 1020 may include data sample component 1025, capability message component 1030, tuning process component 1035, message sending component 1040, first indication receiving component 1045, task execution component 1050, not allowed state receiving component 1055, online tuning process component 1060, or any combination thereof. Communication manager 1020 may be an example of aspects of communication manager 920 as described herein. In some examples, communication manager 1020 or its various components may be configured to perform various operations (e.g., receive, obtain, monitor, output, transmit) using or otherwise in cooperation with receiver 1010, transmitter 1015, or both. For example, communication manager 1020 may receive information from receiver 1010, convey information to transmitter 1015, or integrate with receiver 1010, transmitter 1015, or both to obtain information, output information, or perform various other operations as described herein.

[0211] According to examples disclosed herein, communication manager 1020 may support wireless communication at a UE. Data sample component 1025 may be configured to or otherwise support components for obtaining data samples associated with a first machine learning model related to a task at the UE, wherein a first set of parameters is associated with the first machine learning model. Capability message component 1030 may be configured to or otherwise support components for sending a capability message indicating the UE's ability to perform a tuning process for the first machine learning model. Tuning process component 1035 may be configured to or otherwise support components for performing a tuning process for the first machine learning model to obtain a second set of parameters associated with the first machine learning model based on the UE's ability to perform the tuning process for the first machine learning model. Message sending component 1040 may be configured to or otherwise support components for sending a message indicating at least a portion of the second set of parameters to a network entity based on performing the tuning process for the first machine learning model.

[0212] Additionally or alternatively, according to an example as disclosed herein, the communication manager 1020 may utilize a first set of parameters associated with a first machine learning model for a task to support wireless communication at a UE. The capability message component 1030 may be configured to or otherwise support a component for sending a capability message to a network entity indicating the UE's capability to perform a tuning process of the first machine learning model, wherein the first set of parameters is associated with the first machine learning model. The tuning process component 1035 may be configured to or otherwise support a component for performing a tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model based on the UE's capability to perform the tuning process of the first machine learning model. The message sending component 1040 may be configured to or otherwise support a component for sending a message indicating the second set of parameters to a network entity based on performing the tuning process of the first machine learning model. The first indication receiving component 1045 may be configured to or otherwise support a component for receiving an allowed status indication from a network entity associated with the second set of parameters. The task execution component 1050 may be configured to or otherwise support a component for performing a task using the first set of parameters associated with the first machine learning model or the second set of parameters associated with the first machine learning model based on the received allowed status indication. The not allowed status receiving component 1055 may be configured to or otherwise support a component for receiving a not allowed status indication from a network entity associated with the second set of parameters. The task execution component 1050 may be configured to or otherwise support a component for performing a task using the first set of parameters associated with the first machine learning model based on the received not allowed status indication.

[0213] The task execution component 1050 may be configured to or otherwise support a component for primarily determining whether to use the first set of parameters associated with the first machine learning model or the second set of parameters associated with the first machine learning model to perform a task based at least in part on the received allowed status indication.

[0214] Additionally or alternatively, according to an example as disclosed herein, the communication manager 1020 may utilize a first set of parameters associated with a first machine learning model for a task to support wireless communication at a UE. The capability message component 1030 may be configured to or otherwise support a component for sending a capability message to a network entity indicating the capability of the UE to perform an online tuning process of the first machine learning model, wherein the first set of parameters is associated with the first machine learning model. The first indication receiving component 1045 may be configured to or otherwise support a component for receiving a first indication from a network entity associated with the online tuning process, wherein the first indication includes an activation state or an allowance state. The online tuning process component 1060 may be configured to or otherwise support a component for performing an online tuning process of the first machine learning model based on the capability of the UE to perform a tuning process of the first machine learning model and the received allowance state to obtain a second set of parameters associated with the first machine learning model.

[0215] Figure 11 Block diagram 1100 illustrates a communication manager 1120 that supports model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure. The communication manager 1120 may be an example of aspects of the communication manager 920, the communication manager 1020, or both as described herein. The communication manager 1120 or its various components may be examples of components for performing various aspects of model tuning for cross-node machine learning as described herein. For example, the communication manager 1120 may include a data sample component 1125, a capability message component 1130, a tuning process component 1135, a message sending component 1140, a first indication receiving component 1145, a task execution component 1150, a disallowance state receiving component 1155, an online tuning process component 1160, a machine learning model indication component 1165, a parameter update component 1170, a second indication sending component 1175, a loss function parameter component 1180, an offline tuning process component 1185, an activation request component 1190, a fourth indication receiving component 1195, a forward propagation component 11100, a backpropagation component 11105, an encoder availability component 11110, or any combination thereof. Each of these components may communicate directly or indirectly with each other (e.g., via one or more buses).

[0216] According to examples disclosed herein, communication manager 1120 may support wireless communication at a UE. Data sample component 1125 may be configured to or otherwise support components for obtaining data samples for a first machine learning model associated with a task at the UE, where a first set of parameters is associated with the first machine learning model. Capability message component 1130 may be configured to or otherwise support components for sending a capability message indicating the UE's ability to perform a tuning process of the first machine learning model. Tuning process component 1135 may be configured to or otherwise support components for performing a tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model based on the UE's ability to perform the tuning process of the first machine learning model. Message sending component 1140 may be configured to or otherwise support components for sending a message indicating at least a portion of the second set of parameters to a network entity based on performing the tuning process of the first machine learning model.

[0217] In some examples, the first machine learning model includes an encoder portion of a second machine learning model. In some examples, the third machine learning model includes a decoder portion of the second machine learning model.

[0218] In some examples, to support performing the tuning process of the first machine learning model, loss function parameter component 1180 may be configured to or otherwise support components for receiving a set of parameters associated with a loss function.

[0219] In some examples, forward propagation component 11100 may be configured to or otherwise support components for sending a message associated with a forward propagation process.

[0220] In some examples, backpropagation component 11105 may be configured to or otherwise support components for receiving a message associated with a backpropagation process for adjusting parameters associated with an encoder, where the message indicates a gradient associated with a loss function. In some examples, parameter update component 1170 may be configured to or otherwise support components for updating parameters associated with the encoder based on the message.

[0221] In some examples, loss function parameter component 1180 may be configured to or otherwise support components for receiving a set of parameters associated with a loss function. In some examples, parameter update component 1170 may be configured to or otherwise support components for updating parameters associated with a decoder portion of a second machine learning model based on the set of parameters.

[0222] In some examples, to support sending the capability message, the machine learning model indicating component 1165 may be configured to or otherwise support a component for sending an indication of a set of machine learning models supported by a UE.

[0223] In some examples, the task includes a channel state information (CSI) feedback task.

[0224] In some examples, to support performing a tuning process of a first machine learning model, the parameter update component 1170 may be configured to or otherwise support a component for updating parameters associated with an encoder, parameters associated with a decoder, or both using a second set of parameters.

[0225] In some examples, to support performing a tuning process of a first machine learning model, the online tuning process component 1160 may be configured to or otherwise support a component for performing an online tuning process.

[0226] In some examples, to support performing an online tuning process of a first machine learning model, the parameter update component 1170 may be configured to or otherwise support a component for updating a second set of parameters associated with an encoder using a second set of parameters for the first machine learning model when performing a task.

[0227] In some examples, to support performing a tuning process of a first machine learning model, the offline tuning process component 1185 may be configured to or otherwise support a component for performing an offline tuning process.

[0228] In some examples, to support performing an offline tuning process of a first machine learning model, the parameter update component 1170 may be configured to or otherwise support a component for updating a second set of parameters associated with an encoder using a first set of parameters of the first machine learning model when performing a task.

[0229] In some examples, the encoder availability component 11110 may be configured to or otherwise support a device for sending a third indication to a network entity, where the third indication indicates the availability of a second encoder, and the second encoder is associated with performing an offline tuning process.

[0230] In some examples, to support the execution of the tuning process, the first indication receiving component 1145 may be configured to or otherwise support components for receiving a first indication from a network entity associated with the tuning process of executing a first machine learning model, where the first indication includes an activation state or an allowed state. In some examples, to support the execution of the tuning process, the second indication sending component 1175 may be configured to or otherwise support components for sending a second indication to the network entity in response to the first indication, where the second indication includes an activation indication associated with starting the execution of the tuning process or a deactivation indication associated with stopping the tuning process.

[0231] In some examples, the activation request component 1190 may be configured to or otherwise support components for sending an activation request to the network entity, where receiving the first indication is based on the activation request.

[0232] In some examples, the fourth indication receiving component 1195 may be configured to or otherwise support components for receiving a fourth indication from the network entity, where the fourth indication includes a deactivation indication associated with stopping the tuning process.

[0233] In some examples, to support the tuning process of executing a first machine learning model, the parameter update component 1170 may be configured to or otherwise support components for updating the parameters associated with the encoder, the parameters associated with the decoder, or both using a second parameter set based on the gradients associated with a first parameter set and a second parameter set.

[0234] In some examples, to support updating the parameters associated with the encoder, the parameters associated with the decoder, or both, the loss function parameter component 1180 may be configured to or otherwise support components for receiving a parameter set associated with the loss function. In some examples, to support updating the parameters associated with the encoder, the parameters associated with the decoder, or both, the forward propagation component 11100 may be configured to or otherwise support components for sending a message associated with the forward propagation process.

[0235] In some examples, the backpropagation component 11105 may be configured to or otherwise support components for receiving a message associated with the backpropagation process for adjusting the parameters associated with the encoder, the parameters associated with the decoder, or both, where the message indicates the gradient associated with the loss function. In some examples, the parameter update component 1170 may be configured to or otherwise support components for updating the parameters associated with the encoder, the parameters associated with the decoder, or both based on the message.

[0236] In some examples, the UE receives an indication of the first parameter set via broadcast signaling, dedicated signaling, or both.

[0237] Additionally or alternatively, according to examples disclosed herein, the communication manager 1120 may utilize a first set of parameters associated with a first machine learning model for a task to support wireless communication at the UE. In some examples, the capability message component 1130 may be configured to or otherwise support a component for sending a capability message to a network entity indicating the UE's capability to perform a tuning process of the first machine learning model, wherein the first set of parameters is associated with the first machine learning model. In some examples, the tuning process component 1135 may be configured to or otherwise support a component for performing a tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model based on the UE's capability to perform the tuning process of the first machine learning model. In some examples, the message sending component 1140 may be configured to or otherwise support a component for sending a message indicating the second set of parameters to a network entity based on performing the tuning process of the first machine learning model. The first indication receiving component 1145 may be configured to or otherwise support a component for receiving a permission status indication from a network entity associated with the second set of parameters. The task execution component 1150 may be configured to or otherwise support a component for performing a task using the first set of parameters associated with the first machine learning model or the second set of parameters associated with the first machine learning model based on the received permission status indication. The non-permission status receiving component 1155 may be configured to or otherwise support a component for receiving a non-permission status indication from a network entity associated with the second set of parameters. In some examples, the task execution component 1150 may be configured to or otherwise support a component for performing a task using the first set of parameters associated with the first machine learning model based on the received non-permission status indication.

[0238] Additionally or alternatively, according to examples disclosed herein, the communication manager 1120 may utilize a first set of parameters associated with a first machine learning model for a task to support wireless communication at the UE. In some examples, the capability message component 1130 may be configured to or otherwise support a component for sending a capability message to a network entity indicating the UE's capability to perform an online tuning process of the first machine learning model, wherein the first set of parameters is associated with the first machine learning model. In some examples, the first indication receiving component 1145 may be configured to or otherwise support a component for receiving a first indication from a network entity associated with the online tuning process, wherein the first indication includes an activation status or a permission status. The online tuning process component 1160 may be configured to or otherwise support a component for performing an online tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model based on the UE's capability to perform the tuning process of the first machine learning model and the received permission status.

[0239] Figure 12 FIG. illustrates a system 1200 including a device 1205 that supports model tuning for cross-node machine learning, in accordance with one or more aspects of the present disclosure. The device 1205 may be an example of, or include components of, the device 905, the device 1005, or the UE 115 as described herein. The device 1205 may communicate with one or more network entities 105, one or more UEs 115, or any combination thereof (e.g., wirelessly). The device 1205 may include components for two-way voice and data communication, including components for sending and receiving communications, such as a communication manager 1220, an input / output (I / O) controller 1210, a transceiver 1215, an antenna 1225, a memory 1230, code 1235, and a processor 1240. These components may be electronically communicatively coupled via one or more buses (e.g., bus 1245) or otherwise (e.g., operatively, communicatively, functionally, electronically, electrically).

[0240] The I / O controller 1210 may manage input and output signals of the device 1205. The I / O controller 1210 may also manage peripheral devices not integrated into the device 1205. In some cases, the I / O controller 1210 may represent a physical connection or port to an external peripheral device. In some cases, the I / O controller 1210 may utilize an operating system, such as iOS ® , ANDROID ® , MS-DOS ® , MS-WINDOWS ® , OS / 2 ® , UNIX ® , LINUX ® or another known operating system. Additionally or alternatively, the I / O controller 1210 may represent, or interact with, a modem, a keyboard, a mouse, a touch screen, or similar device. In some cases, the I / O controller 1210 may be implemented as part of a processor (such as the processor 1240). In some cases, a user may interact with the device 1205 via the I / O controller 1210 or via hardware components controlled by the I / O controller 1210.

[0241] In some cases, device 1205 may include a single antenna 1225. However, in some other cases, device 1205 may have more than one antenna 1225, and the more than one antenna may be capable of concurrently transmitting or receiving multiple wireless transmissions. Transceiver 1215 may communicate bidirectionally via one or more antennas 1225, wired or wireless links as described herein. For example, transceiver 1215 may represent a wireless transceiver and may communicate bidirectionally with another wireless transceiver. Transceiver 1215 may also include a modem for: modulating a packet; providing the modulated packet to one or more antennas 1225 for transmission; and demodulating a packet received from one or more antennas 1225. Transceiver 1215 or transceiver 1215 and one or more antennas 1225 may be examples of transmitter 915, transmitter 1015, receiver 910, receiver 1010 or any combination thereof or components thereof as described herein.

[0242] Memory 1230 may include random access memory (RAM) and read only memory (ROM). Memory 1230 may store computer-readable, computer-executable code 1235 including instructions that, when executed by processor 1240, cause device 1205 to perform the various functions described herein. Code 1235 may be stored in a non-transitory computer-readable medium (such as system memory or another type of memory). In some cases, code 1235 may not be directly executable by processor 1240 but may (e.g., when compiled and executed) cause a computer to perform the functions described herein. In some cases, memory 1230 may contain a basic input / output system (BIOS) and the like, which may control basic hardware or software operations, such as interactions with peripheral components or devices.

[0243] Processor 1240 may include intelligent hardware devices (e.g., general purpose processor, DSP, CPU, microcontroller, ASIC, FPGA, programmable logic device, discrete gate or transistor logic, discrete hardware component, or any combination thereof). In some cases, processor 1240 may be configured to operate a memory array using a memory controller. In some other cases, the memory controller may be integrated into processor 1240. Processor 1240 may be configured to execute computer-readable instructions stored in a memory (e.g., memory 1230) to cause device 1205 to perform various functions (e.g., support functions or tasks for model tuning for cross-node machine learning). For example, device 1205 or components of device 1205 may include processor 1240 and memory 1230 coupled to or coupled with processor 1240, and processor 1240 and memory 1230 are configured to perform the various functions described herein.

[0244] According to examples disclosed herein, communication manager 1220 may support wireless communication at a UE. For example, communication manager 1220 may be configured to or otherwise support components for obtaining data samples for a first machine learning model associated with a task at the UE, where a first set of parameters is associated with the first machine learning model. Communication manager 1220 may be configured to or otherwise support components for sending a capability message indicating the UE's ability to perform a tuning process of the first machine learning model. Communication manager 1220 may be configured to or otherwise support components for performing a tuning process of the first machine learning model based on the UE's ability to perform the tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model. Communication manager 1220 may be configured to or otherwise support components for sending a message indicating at least a portion of the second set of parameters to a network entity based on performing the tuning process of the first machine learning model.

[0245] Additionally or alternatively, according to examples disclosed herein, communication manager 1220 may support wireless communication at a UE using a first set of parameters associated with a first machine learning model for a task. For example, communication manager 1220 may be configured to or otherwise support components for sending a capability message indicating the UE's ability to perform a tuning process of the first machine learning model, where the first set of parameters is associated with the first machine learning model. Communication manager 1220 may be configured to or otherwise support components for performing a tuning process of the first machine learning model based on the UE's ability to perform the tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model. Communication manager 1220 may be configured to or otherwise support components for sending a message indicating the second set of parameters to a network entity based on performing the tuning process of the first machine learning model. Communication manager 1220 may be configured to or otherwise support components for receiving an allowed status indication from a network entity associated with the second set of parameters. Communication manager 1220 may be configured to or otherwise support components for performing the task using the first set of parameters associated with the first machine learning model or the second set of parameters associated with the first machine learning model based on the received allowed status indication. Communication manager 1220 may be configured to or otherwise support components for receiving a not-allowed status indication from a network entity associated with the second set of parameters. Communication manager 1220 may be configured to or otherwise support components for performing the task using the first set of parameters associated with the first machine learning model based on the received not-allowed status indication.

[0246] Additionally or alternatively, according to examples as disclosed herein, the communication manager 1220 may utilize a first set of parameters associated with a first machine learning model for a task to support wireless communication at the UE. For example, the communication manager 1220 may be configured as or otherwise support a component for sending a capability message to a network entity indicating the UE's ability to perform an online tuning process of the first machine learning model, where the first set of parameters is associated with the first machine learning model. The communication manager 1220 may be configured as or otherwise support a component for receiving a first indication from a network entity associated with the online tuning process, where the first indication includes an activation state or an allowance state. The communication manager 1220 may be configured as or otherwise support a component for performing an online tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model based on the UE's ability to perform the tuning process of the first machine learning model and the received allowance state.

[0247] By including or configuring the communication manager 1220 according to examples as described herein, the device 1205 may support techniques for reducing latency and more efficiently utilizing communication resources. For example, by sending a capability message to a network entity, the UE may notify the network entity of its ability to autonomously tune a second machine learning model. Autonomously performing the tuning process may cause the processor of the device 1205 to more efficiently tune the second machine learning model and reduce latency in communications using the second machine learning model.

[0248] In some examples, the communication manager 1220 may be configured to use or otherwise cooperate with the transceiver 1215, one or more antennas 1225, or any combination thereof to perform various operations (e.g., receive, monitor, transmit). Although the communication manager 1220 is illustrated as a separate component, in some examples, one or more functions described with reference to the communication manager 1220 may be supported or performed by the processor 1240, the memory 1230, the code 1235, or any combination thereof. For example, the code 1235 may include instructions that can be executed by the processor 1240 to cause the device 1205 to perform various aspects of model tuning for cross-node machine learning as described herein, or the processor 1240 and the memory 1230 may otherwise be configured to execute or support such operations.

[0249] Figure 13FIG. 1300 is a block diagram illustrating a device 1305 that supports model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure. The device 1305 may be an example of aspects of the network entity 105 described herein. The device 1305 may include a receiver 1310, a transmitter 1315, and a communication manager 1320. The device 1305 may also include a processor. Each of these components may communicate with each other (e.g., via one or more buses).

[0250] The receiver 1310 may provide components for obtaining (e.g., receiving, determining, identifying) information associated with various channels (e.g., control channels, data channels, information channels, channels associated with a protocol stack), such as user data, control information, or any combination thereof (e.g., I / Q samples, symbols, packets, protocol data units, service data units). The information may be delivered to other components of the device 1305. In some examples, the receiver 1310 may support obtaining information by receiving signals via one or more antennas. Additionally or alternatively, the receiver 1310 may support obtaining information by receiving signals via one or more wired (e.g., electrical, fiber optic) interfaces, wireless interfaces, or any combination thereof.

[0251] The transmitter 1315 may provide components for outputting (e.g., transmitting, providing, conveying, delivering) information generated by other components of the device 1305. For example, the transmitter 1315 may output information associated with various channels (e.g., control channels, data channels, information channels, channels associated with a protocol stack), such as user data, control information, or any combination thereof (e.g., I / Q samples, symbols, packets, protocol data units, service data units). In some examples, the transmitter 1315 may support outputting information by transmitting signals via one or more antennas. Additionally or alternatively, the transmitter 1315 may support outputting information by transmitting signals via one or more wired (e.g., electrical, fiber optic) interfaces, wireless interfaces, or any combination thereof. In some examples, the transmitter 1315 and the receiver 1310 may be co-located in a transceiver that may include a modem or be coupled to a modem.

[0252] The communication manager 1320, the receiver 1310, the transmitter 1315, or various combinations thereof or their various components may be examples of components for performing various aspects of model tuning for cross-node machine learning as described herein. For example, the communication manager 1320, the receiver 1310, the transmitter 1315, or various combinations thereof or components may support methods for performing one or more of the functions described herein.

[0253] In some examples, the communication manager 1320, the receiver 1310, the transmitter 1315, or various combinations or components thereof may be implemented in hardware (e.g., in a communication management circuit). The hardware may include a processor, a DSP, a CPU, an ASIC, an FPGA, or other programmable logic device, a microcontroller, discrete gate or transistor logic, discrete hardware components, or any combination thereof configured to or otherwise supporting components for performing the functions described in this disclosure. In some examples, a processor and a memory coupled to the processor may be configured to perform one or more of the functions described herein (e.g., by the processor executing instructions stored in the memory).

[0254] Additionally or alternatively, in some examples, the communication manager 1320, the receiver 1310, the transmitter 1315, or various combinations or components thereof may be implemented in code executed by a processor (e.g., implemented as communication management software or firmware). If implemented in code executed by a processor, the functions of the communication manager 1320, the receiver 1310, the transmitter 1315, or various combinations or components thereof may be performed by a general-purpose processor, a DSP, a CPU, an ASIC, an FPGA, a microcontroller, or any combination of these or other programmable logic devices configured to or otherwise supporting components for performing the functions described in this disclosure.

[0255] In some examples, the communication manager 1320 may be configured to use or otherwise cooperate with the receiver 1310, the transmitter 1315, or both to perform various operations (e.g., receive, obtain, monitor, output, transmit). For example, the communication manager 1320 may receive information from the receiver 1310, convey information to the transmitter 1315, or integrate in combination with the receiver 1310, the transmitter 1315, or both to obtain information, output information, or perform various other operations as described herein.

[0256] According to examples disclosed herein, the communication manager 1320 may support wireless communication at a network entity. For example, the communication manager 1320 may be a component configured to or otherwise supporting receiving, from a UE, a capability message indicating the UE's ability to perform a tuning process of a first machine learning model associated with a first set of parameters. The communication manager 1320 may be a component configured to or otherwise supporting receiving, from a UE, a message indicating at least a portion of a second set of parameters.

[0257] By including or configuring a communication manager 1320 according to examples as described herein, a device 1305 (e.g., a control receiver 1310, a transmitter 1315, a communication manager 1320, or a processor combined with or otherwise coupled to them) can support techniques for reducing power consumption and more efficiently utilizing communication resources. For example, by sending a capabilities message to a network entity, a UE can notify the network entity of the ability to autonomously tune a second machine learning model. Autonomously performing the tuning process can cause the processor of the device 1305 to more efficiently tune the second machine learning model and reduce latency in communications using the second machine learning model.

[0258] Figure 14 Block diagram 1400 illustrates a device 1405 that supports model tuning for cross-node machine learning according to one or more aspects of the present disclosure. The device 1405 can be an example of aspects of the device 1305 or the network entity 105 as described herein. The device 1405 can include a receiver 1410, a transmitter 1415, and a communication manager 1420. The device 1405 can also include a processor. Each of these components can communicate with each other (e.g., via one or more buses).

[0259] The receiver 1410 can provide components for obtaining (e.g., receiving, determining, identifying) information associated with various channels (e.g., control channels, data channels, information channels, channels associated with a protocol stack) such as user data, control information, or any combination thereof (e.g., I / Q samples, symbols, packets, protocol data units, service data units). The information can be delivered to other components of the device 1405. In some examples, the receiver 1410 can support obtaining information by receiving signals via one or more antennas. Additionally or alternatively, the receiver 1410 can support obtaining information by receiving signals via one or more wired (e.g., electrical, fiber optic) interfaces, wireless interfaces, or any combination thereof.

[0260] The transmitter 1415 may provide means for outputting (e.g., transmitting, providing, conveying, transmitting) information generated by other components of the device 1405. For example, the transmitter 1415 may output information associated with various channels (e.g., control channels, data channels, information channels, channels associated with a protocol stack), such as user data, control information, or any combination thereof (e.g., I / Q samples, symbols, packets, protocol data units, service data units). In some examples, the transmitter 1415 may support outputting information by transmitting signals via one or more antennas. Additionally or alternatively, the transmitter 1415 may support outputting information by transmitting signals via one or more wired (e.g., electrical, optical) interfaces, wireless interfaces, or any combination thereof. In some examples, the transmitter 1415 and the receiver 1410 may be co-located in a transceiver, which may include a modem or be coupled to a modem.

[0261] Device 1405 or its various components may be examples of components for performing various aspects of model tuning for cross-node machine learning as described herein. For example, communication manager 1420 may include capability receiving component 1425, message receiving component 1430, or any combination thereof. Communication manager 1420 may be an example of various aspects of communication manager 1320 as described herein. In some examples, communication manager 1420 or its various components may be configured to perform various operations (e.g., receive, obtain, monitor, output, send) using or otherwise in conjunction with receiver 1410, transmitter 1415, or both. For example, communication manager 1420 may receive information from receiver 1410, transmit information to transmitter 1415, or integrate with receiver 1410, transmitter 1415, or both in combination to obtain information, output information, or perform various other operations as described herein.

[0262] According to examples as disclosed herein, the communication manager 1420 may support wireless communications at a network entity. The capability receiving component 1425 may be configured to or otherwise support means for receiving a capability message from a UE indicating the capability of the UE to perform a tuning process of a first machine learning model associated with a first set of parameters at the UE. The message receiving component 1430 may be configured to or otherwise support means for receiving a message from the UE indicating at least a portion of a second set of parameters.

[0263] Figure 15FIG. 1500 is a block diagram illustrating a communication manager 1520 that supports model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure. The communication manager 1520 may be an example of aspects of the communication manager 1320, the communication manager 1420, or both as described herein. The communication manager 1520 or its various components may be examples of components for performing various aspects of model tuning for cross-node machine learning as described herein. For example, the communication manager 1520 may include a capabilities receiving component 1525, a message receiving component 1530, a machine learning model indication receiving component 1535, a first indication sending component 1540, a second indication receiving component 1545, an activation request receiving component 1550, a third indication sending component 1555, or any combination thereof. Each of these components may communicate directly or indirectly with each other (e.g., via one or more buses), which may include communication within protocol layers of a protocol stack, communication associated with logical channels of the protocol stack (e.g., between protocol layers of the protocol stack, within devices, components, or virtualized components associated with the network entity 105, between devices, components, or virtualized components associated with the network entity 105), or any combination thereof.

[0264] In accordance with examples as disclosed herein, the communication manager 1520 may support wireless communication at a network entity. The capabilities receiving component 1525 may be a component configured to or otherwise support receiving a capabilities message from a UE indicating the UE's capabilities to perform a tuning process for a first machine learning model associated with a first set of parameters. The message receiving component 1530 may be a component configured to or otherwise support receiving a message from the UE indicating at least a portion of a second set of parameters.

[0265] In some examples, the first machine learning model includes an encoder portion of a second machine learning model. In some examples, the third machine learning model includes a decoder portion of the second machine learning model.

[0266] In some examples, to support receiving the capabilities message, the machine learning model indication receiving component 1535 may be a component configured to or otherwise support receiving an indication of a set of machine learning models supported by the UE.

[0267] In some examples, receiving the message is associated with receiving channel state information feedback.

[0268] In some examples, the first indication sending component 1540 may be configured to or otherwise support components for sending a first indication associated with performing a tuning process of a first machine learning model to a UE, where the first indication includes an activation state or an allowed state. In some examples, the second indication receiving component 1545 may be configured to or otherwise support components for receiving a second indication from the UE in response to the first indication, where the second indication includes an activation indication associated with starting to perform the tuning process or a deactivation indication associated with stopping the tuning process.

[0269] In some examples, the activation request receiving component 1550 may be configured to or otherwise support components for receiving an activation request from the UE, where sending the first indication is based on the activation request.

[0270] In some examples, the third indication sending component 1555 may be configured to or otherwise support components for sending a third indication to the UE, where the third indication includes a deactivation indication associated with stopping the tuning process.

[0271] Figure 16 FIG. illustrates a system 1600 including a device 1605 that supports model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure. The device 1605 may be an example of the device 1305, the device 1405, or the network entity 105 as described herein, or include components thereof. The device 1605 may communicate with one or more network entities 105, one or more UEs 115, or any combination thereof, and such communication may include communication via one or more wired interfaces, via one or more wireless interfaces, or any combination thereof. The device 1605 may include components that support outputting and obtaining communication, such as a communication manager 1620, a transceiver 1610, an antenna 1615, a memory 1625, code 1630, and a processor 1635. These components may be electronically communicated or otherwise (e.g., operatively, communicatively, functionally, electronically, electrically) coupled via one or more buses (e.g., bus 1640).

[0272] The transceiver 1610 may support bidirectional communication via a wired link, a wireless link, or both as described herein. In some examples, the transceiver 1610 may include a wired transceiver and may communicate bidirectionally with another wired transceiver. Additionally or alternatively, in some examples, the transceiver 1610 may include a wireless transceiver and may communicate bidirectionally with another wireless transceiver. In some examples, the device 1605 may include one or more antennas 1615, which may be capable of (e.g., concurrently) sending or receiving wireless transmissions. The transceiver 1610 may also include a modem for: modulating a signal; providing the modulated signal for transmission (e.g., via one or more antennas 1615, via a wired transmitter); receiving the modulated signal (e.g., from one or more antennas 1615, from a wired receiver); and demodulating the signal. In some implementations, the transceiver 1610 may include one or more interfaces, such as one or more interfaces coupled to one or more antennas 1615 configured to support various receiving or obtaining operations, or one or more interfaces coupled to one or more antennas 1615 configured to support various transmitting or outputting operations, or a combination thereof. In some implementations, the transceiver 1610 may include or be configured to be coupled to one or more processors or memory components, which are operable to perform or support operations based on received or obtained information or signals, or to generate information or other signals for transmission or other output, or any combination of the above. In some implementations, the transceiver 1610, or the transceiver 1610 and one or more antennas 1615, or the transceiver 1610 and one or more antennas 1615 and one or more processors or memory components (e.g., processor 1635 or memory 1625 or both) may be included in a chip or chip assembly installed in the device 1605. In some examples, the transceiver may be operable to support communications via one or more communication links (eg, communication link 125 , backhaul communication link 120 , midhaul communication link 162 , fronthaul communication link 168 ).

[0273] The memory 1625 may include RAM and ROM. The memory 1625 may store computer-readable, computer-executable code 1630 including instructions that, when executed by the processor 1635, cause the device 1605 to perform the various functions described herein. The code 1630 may be stored in a non-transitory computer-readable medium (such as system memory or another type of memory). In some cases, the code 1630 may not be directly executable by the processor 1635 but may (e.g., when compiled and executed) cause a computer to perform the functions described herein. In some cases, the memory 1625 may contain BIOS and the like, which may control basic hardware or software operations, such as interactions with peripheral components or devices.

[0274] Processor 1635 may include intelligent hardware devices (e.g., general-purpose processors, DSPs, ASICs, CPUs, FPGAs, microcontrollers, programmable logic devices, discrete gate or transistor logic, discrete hardware components, or any combination thereof). In some cases, processor 1635 may be configured to operate a memory array using a memory controller. In some other cases, the memory controller may be integrated into processor 1635. Processor 1635 may be configured to execute computer-readable instructions stored in a memory (e.g., memory 1625) to cause device 1605 to perform various functions (e.g., support various functions or tasks for model tuning for cross-node machine learning). For example, device 1605 or components of device 1605 may include processor 1635 and memory 1625 coupled to processor 1635, and processor 1635 and memory 1625 are configured to perform the various functions described herein. Processor 1635 may be an example of a cloud computing platform (e.g., one or more physical nodes and supporting software such as an operating system, virtual machine, or container instance) that may host functions (e.g., by executing code 1630) to perform the functions of device 1605. Processor 1635 may be any one or more suitable processors capable of executing scripts or instructions of one or more software programs stored in device 1605 (such as within memory 1625). In some specific implementations, processor 1635 may be a component of a processing system. A processing system generally may refer to a system or a collection of machines or components that receive inputs and process these inputs to produce outputs (which may be delivered to other systems or components of, for example, device 1605). For example, the processing system of device 1605 may refer to a system that includes various other components or sub-components of device 1605, such as processor 1635, or transceiver 1610, or communication manager 1620, or a combination of other components or components of device 1605. The processing system of device 1605 may interface with other components of device 1605 and may process information (such as inputs or signals) received from other components or output information to other components. For example, a chip or modem of device 1605 may include a processing system and one or more interfaces for outputting information or for obtaining information or both. One or more interfaces may be implemented as or otherwise include a first interface configured to output information and a second interface configured to obtain information or the same interface configured to output information and obtain information, and so on. In some specific implementations, one or more interfaces may refer to an interface between the processing system of a chip or modem and a transmitter such that device 1605 may transmit information output from the chip or modem.Additionally or alternatively, in some embodiments, one or more interfaces may refer to an interface between a processing system of a chip or modem and a receiver, such that the device 1605 can obtain information or signal inputs, and this information can be delivered to the processing system. Those of ordinary skill in the art will readily recognize that the first interface can also obtain information or signal inputs, and the second interface can also output information or signal outputs.

[0275] In some examples, the bus 1640 may support communication within a protocol layer of a protocol stack (e.g., within the protocol layer). In some examples, the bus 1640 may support communication associated with a logical channel of a protocol stack (e.g., between protocol layers of the protocol stack), which may include communication performed within components of the device 1605 or between different components of the device 1605 that may be co-located or located at different positions (e.g., where the device 1605 may refer to a system in which one or more of the communication manager 1620, transceiver 1610, memory 1625, code 1630, and processor 1635 may be located in one component or divided among different components).

[0276] In some examples, the communication manager 1620 may manage aspects of communication with the core network 130 (e.g., via one or more wired or wireless backhaul links). For example, the communication manager 1620 may manage the delivery of data communication for client devices such as one or more UEs 115. In some examples, the communication manager 1620 may manage communication with other network entities 105 and may include a controller or scheduler for coordinating with other network entities 105 to control communication with the UEs 115. In some examples, the communication manager 1620 may support the X2 interface within the LTE / LTE-A wireless communication network technology to provide communication between network entities 105.

[0277] According to examples disclosed herein, the communication manager 1620 may support wireless communication at a network entity. For example, the communication manager 1620 may be configured or otherwise support components for receiving from a UE a capability message indicating the UE's ability to perform a tuning process of a first machine learning model associated with a first set of parameters. The communication manager 1620 may be configured or otherwise support components for receiving from a UE a message indicating at least a portion of a second set of parameters.

[0278] By or configured according to examples as described herein including a communication manager 1620, the device 1605 may support techniques for reducing latency and more efficiently utilizing communication resources. For example, by sending a capabilities message to a network entity, the UE may notify the network entity of the ability to autonomously tune a second machine learning model. Autonomously performing the tuning process may cause the processor of the device 1605 to more efficiently tune the second machine learning model and reduce latency in communications using the second machine learning model.

[0279] In some examples, the communication manager 1620 may be configured to perform various operations (e.g., receive, obtain, monitor, output, transmit) using or otherwise in conjunction with the transceiver 1610, one or more antennas 1615 (e.g., where applicable), or any combination thereof. Although the communication manager 1620 is illustrated as a separate component, in some examples, one or more functions described with reference to the communication manager 1620 may be supported or performed by the transceiver 1610, the processor 1635, the memory 1625, the code 1630, or any combination thereof. For example, the code 1630 may include instructions that can be executed by the processor 1635 to cause the device 1605 to perform various aspects of model tuning for cross-node machine learning as described herein, or the processor 1635 and the memory 1625 may otherwise be configured to perform or support such operations.

[0280] Figure 17 A flowchart illustrating a method 1700 that supports model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure is shown. The operations of method 1700 may be implemented by a UE or components thereof as described herein. For example, the operations of method 1700 may be performed by a UE 115 as described with reference to Figures 1 to 12 In some examples, the UE may execute an instruction set to control functional elements of the UE to perform the described functions. Additionally or alternatively, the UE may use dedicated hardware to perform aspects of the described functions.

[0281] At 1705, the method may include: obtaining a data sample associated with a task at the UE, where a first set of parameters is associated with the first machine learning model. The operation of 1705 may be performed in accordance with examples as disclosed herein. In some examples, aspects of the operation of 1705 may be performed by a data sample component 1125 as described with reference to Figure 11 At 1710, the method may include: sending a capabilities message indicating the ability of the UE to perform a tuning process of the first machine learning model. The operation of 1710 may be performed in accordance with examples as disclosed herein. In some examples, aspects of the operation of 1710 may be performed by

[0282] as described with reference toFigure 11 Execute by the described capability message component 1130.

[0283] At 1715, the method may include: Executing the tuning process of the first machine learning model based on the capability of the UE to execute the tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model. The operation at 1715 may be performed according to the examples disclosed herein. In some examples, aspects of the operation at 1715 may be performed by the tuning process component 1135 as described with reference to Figure 11 Execute by the described tuning process component 1135.

[0284] At 1720, the method may include: Sending a message indicating at least a portion of the second set of parameters to a network entity based on executing the tuning process of the first machine learning model. The operation at 1720 may be performed according to the examples disclosed herein. In some examples, aspects of the operation at 1720 may be performed by the message sending component 1140 as described with reference to Figure 11 Execute by the described message sending component 1140.

[0285] Figure 18 A flowchart illustrating a method 1800 that supports model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure is shown. The operations of method 1800 may be implemented by a UE or its components as described herein. For example, the operations of method 1800 may be performed by the UE 115 as described with reference to Figures 1 to 12 Execute by the described UE 115. In some examples, the UE may execute an instruction set to control the functional elements of the UE to perform the described functions. Additionally or alternatively, the UE may use dedicated hardware to perform aspects of the described functions.

[0286] At 1805, the method may include: Obtaining a data sample of a first machine learning model associated with a task at the UE, wherein a first set of parameters is associated with the first machine learning model. The operation at 1805 may be performed according to the examples disclosed herein. In some examples, aspects of the operation at 1805 may be performed by the data sample component 1125 as described with reference to Figure 11 Execute by the described data sample component 1125.

[0287] At 1810, the method may include: Sending a capability message indicating the capability of the UE to execute the tuning process of the first machine learning model. The operation at 1810 may be performed according to the examples disclosed herein. In some examples, aspects of the operation at 1810 may be performed by the capability message component 1130 as described with reference to Figure 11 Execute by the described capability message component 1130.

[0288] At 1815, the method may include: receiving a set of parameters associated with a loss function. The operation at 1815 may be performed according to the examples disclosed herein. In some examples, aspects of the operation at 1815 may be performed by a loss function parameter component 1180 as described with reference to Figure 11 The loss function parameter component 1180 described.

[0289] At 1820, the method may include: performing a tuning process of the first machine learning model based on the ability of the UE to perform the tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model. The operation at 1820 may be performed according to the examples disclosed herein. In some examples, aspects of the operation at 1820 may be performed by a tuning process component 1135 as described with reference to Figure 11 The tuning process component 1135 described.

[0290] At 1825, the method may include: sending a message indicating at least a portion of the second set of parameters to a network entity based on performing the tuning process of the first machine learning model. The operation at 1825 may be performed according to the examples disclosed herein. In some examples, aspects of the operation at 1825 may be performed by a message sending component 1140 as described with reference to Figure 11 The message sending component 1140 described.

[0291] Figure 19 FIG. shows a flow chart of a method 1900 supporting model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure. The operations of method 1900 may be implemented by a UE or its components as described herein. For example, the operations of method 1900 may be performed by a UE 115 as described with reference to Figures 1 to 12 The UE 115 described. In some examples, the UE may execute an instruction set to control functional elements of the UE to perform the described functions. Additionally or alternatively, the UE may use dedicated hardware to perform aspects of the described functions.

[0292] At 1905, the method may include: sending a capability message to a network entity indicating the ability of the UE to perform a tuning process of a first machine learning model, where a first set of parameters is associated with the first machine learning model. The operation at 1905 may be performed according to the examples disclosed herein. In some examples, aspects of the operation at 1905 may be performed by a capability message component 1130 as described with reference to Figure 11 The capability message component 1130 described.

[0293] At 1910, the method may include: performing the tuning process of the first machine learning model based on the UE's ability to perform the tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model. The operation at 1910 may be performed according to the examples disclosed herein. In some examples, aspects of the operation at 1910 may be performed by the tuning process component 1135 as described with reference to Figure 11 as described.

[0294] At 1915, the method may include: sending a message indicating the second set of parameters to a network entity based on performing the tuning process of the first machine learning model. The operation at 1915 may be performed according to the examples disclosed herein. In some examples, aspects of the operation at 1915 may be performed by the message sending component 1140 as described with reference to Figure 11 as described.

[0295] At 1920, the method may include: receiving an allowed status indication from the network entity associated with the second set of parameters. The operation at 1920 may be performed according to the examples disclosed herein. In some examples, aspects of the operation at 1920 may be performed by the first indication receiving component 1145 as described with reference to Figure 11 as described.

[0296] At 1925, the method may include: performing the task using the first set of parameters associated with the first machine learning model or the second set of parameters associated with the first machine learning model based on the received allowed status indication. The operation at 1925 may be performed according to the examples disclosed herein. In some examples, aspects of the operation at 1925 may be performed by the task execution component 1150 as described with reference to Figure 11 as described.

[0297] At 1930, the method may include: receiving a not allowed status indication from the network entity associated with the second set of parameters. The operation at 1930 may be performed according to the examples disclosed herein. In some examples, aspects of the operation at 1930 may be performed by the not allowed status receiving component 1155 as described with reference to Figure 11 as described.

[0298] At 1935, the method may include: performing the task using the first set of parameters associated with the first machine learning model based on the received not allowed status indication. The operation at 1935 may be performed according to the examples disclosed herein. In some examples, aspects of the operation at 1935 may be performed by the task execution component 1150 as described with reference to Figure 11 as described.

[0299] Figure 20FIG. 2000 is a flow chart illustrating a method 2000 that supports model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure. Operations of method 2000 may be implemented by a UE or components thereof as described herein. For example, operations of method 2000 may be performed by a UE 115 as described with reference to Figures 1 to 12 In some examples, the UE may execute an instruction set to control functional elements of the UE to perform the described functions. Additionally or alternatively, the UE may use dedicated hardware to perform aspects of the described functions.

[0300] At 2005, the method may include: sending a capability message to a network entity indicating the UE's ability to perform an online tuning process of a first machine learning model, where a first set of parameters is associated with the first machine learning model. The operation of 2005 may be performed in accordance with examples as disclosed herein. In some examples, aspects of the operation of 2005 may be performed by a capability message component 1130 as described with reference to Figure 11 In some examples, the UE may execute an instruction set to control functional elements of the UE to perform the described functions. Additionally or alternatively, the UE may use dedicated hardware to perform aspects of the described functions.

[0301] At 2010, the method may include: receiving a first indication from the network entity associated with the online tuning process, where the first indication includes an activation state or a permission state. The operation of 2010 may be performed in accordance with examples as disclosed herein. In some examples, aspects of the operation of 2010 may be performed by a first indication receiving component 1145 as described with reference to Figure 11 In some examples, the UE may execute an instruction set to control functional elements of the UE to perform the described functions. Additionally or alternatively, the UE may use dedicated hardware to perform aspects of the described functions.

[0302] At 2015, the method may include: performing the online tuning process of the first machine learning model based on the UE's ability to perform the tuning process of the first machine learning model and the received permission state to obtain a second set of parameters associated with the first machine learning model. The operation of 2015 may be performed in accordance with examples as disclosed herein. In some examples, aspects of the operation of 2015 may be performed by an online tuning process component 1160 as described with reference to Figure 11 In some examples, the UE may execute an instruction set to control functional elements of the UE to perform the described functions. Additionally or alternatively, the UE may use dedicated hardware to perform aspects of the described functions.

[0303] Figure 21 FIG. 2100 is a flow chart illustrating a method 2100 that supports model tuning for cross-node machine learning in accordance with one or more aspects of the present disclosure. Operations of method 2100 may be implemented by a network entity or components thereof as described herein. For example, operations of method 2100 may be performed by a network entity as described with reference to Figures 1 to 8 as well as Figures 13 to 16 In some examples, the network entity may execute an instruction set to control functional elements of the network entity to perform the described functions. Additionally or alternatively, the network entity may use dedicated hardware to perform aspects of the described functions.

[0304] At 2105, the method may include: receiving, from a UE, a capability message indicating the UE's ability to perform a tuning process of a first machine learning model associated with a first set of parameters. The operation at 2105 may be performed according to the examples disclosed herein. In some examples, aspects of the operation at 2105 may be performed by a capability receiving component 1525 as described with reference to Figure 15 The capability receiving component 1525 described.

[0305] At 2110, the method may include: receiving, from the UE, a message indicating at least a portion of a second set of parameters. The operation at 2110 may be performed according to the examples disclosed herein. In some examples, aspects of the operation at 2110 may be performed by a message receiving component 1530 as described with reference to Figure 15 The message receiving component 1530 described.

[0306] An overview of aspects of the present disclosure is provided below:

[0307] Aspect 1: A method for wireless communication at a UE, the method including: obtaining data samples associated with a task at the UE, wherein a first set of parameters is associated with the first machine learning model; sending a capability message indicating the UE's ability to perform a tuning process of the first machine learning model; performing the tuning process of the first machine learning model at least in part based on the UE's ability to perform the tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model; and sending a message indicating at least a portion of the second set of parameters to a network entity at least in part based on performing the tuning process of the first machine learning model.

[0308] Aspect 2: The method according to aspect 1, wherein the first machine learning model includes an encoder portion of a second machine learning model, and a third machine learning model includes a decoder portion of the second machine learning model.

[0309] Aspect 3: The method according to aspect 2, wherein performing the tuning process of the first machine learning model further includes: receiving a set of parameters associated with a loss function.

[0310] Aspect 4: The method according to aspect 3, the method further including: sending a message associated with a forward propagation process.

[0311] Aspect 5: The method according to any one of aspects 3 to 4, the method further including: receiving a message associated with a backpropagation process for adjusting parameters associated with an encoder, wherein the message indicates a gradient associated with the loss function; and updating the parameters associated with the encoder at least in part based on the message.

[0312] Aspect 6: The method according to any one of Aspects 3 to 5, the method further comprising: receiving the set of parameters associated with the loss function; and updating, at least in part based on the set of parameters, the parameters associated with the decoder portion of the second machine learning model.

[0313] Aspect 7: The method according to any one of Aspects 1 to 6, wherein sending the capability message comprises: sending an indication of the set of machine learning models supported by the UE.

[0314] Aspect 8: The method according to any one of Aspects 1 to 7, wherein the task comprises a CSI feedback task.

[0315] Aspect 9: The method according to any one of Aspects 1 to 8, wherein performing the tuning process of the first machine learning model further comprises: updating the parameters associated with the encoder, the parameters associated with the decoder, or both, using the second set of parameters.

[0316] Aspect 10: The method according to Aspect 9, wherein performing the tuning process of the first machine learning model comprises: performing an online tuning process.

[0317] Aspect 11: The method according to Aspect 10, wherein performing the online tuning process of the first machine learning model comprises: when performing the task, updating the second set of parameters associated with the encoder using the second set of parameters for the first machine learning model.

[0318] Aspect 12: The method according to any one of Aspects 9 to 11, wherein performing the tuning process of the first machine learning model comprises: performing an offline tuning process.

[0319] Aspect 13: The method according to Aspect 12, wherein performing the offline tuning process of the first machine learning model comprises: when performing the task, updating the second set of parameters associated with the encoder using the first set of parameters for the first machine learning model.

[0320] Aspect 14: The method according to any one of Aspects 12 to 13, the method further comprising: sending a third indication to the network entity, wherein the third indication indicates the availability of a second encoder, wherein the second encoder is associated with performing the offline tuning process.

[0321] Aspect 15: The method according to any one of Aspects 1 to 14, wherein performing the tuning process further comprises: receiving a first indication from the network entity associated with performing the tuning process of the first machine learning model, wherein the first indication includes an activation state or a permission state; and sending a second indication to the network entity in response to the first indication, wherein the second indication includes an activation indication associated with starting to perform the tuning process or a deactivation indication associated with stopping the tuning process.

[0322] Aspect 16: The method according to Aspect 15, the method further comprising: sending an activation request to the network entity, wherein receiving the first indication is at least partially based on the activation request.

[0323] Aspect 17: The method according to any one of Aspects 15 to 16, the method further comprising: receiving a fourth indication from the network entity, wherein the fourth indication includes a deactivation indication associated with stopping the tuning process.

[0324] Aspect 18: The method according to any one of Aspects 1 to 17, wherein performing the tuning process of the first machine learning model further comprises: using the second parameter set to update the parameters associated with the encoder, the parameters associated with the decoder, or both, at least partially based on the gradients associated with the first parameter set and the second parameter set.

[0325] Aspect 19: The method according to Aspect 18, wherein updating the parameters associated with the encoder, the parameters associated with the decoder, or both further comprises: receiving a parameter set associated with a loss function; and sending a message associated with the forward propagation process.

[0326] Aspect 20: The method according to Aspect 19, the method further comprising: receiving a message associated with a backpropagation process for adjusting the parameters associated with the encoder, the parameters associated with the decoder, or both, wherein the message indicates the gradient associated with the loss function; and updating the parameters associated with the encoder, the parameters associated with the decoder, or both at least partially based on the message.

[0327] Aspect 21: The method according to any one of Aspects 1 to 20, wherein the UE receives the indication of the first parameter set via broadcast signaling, dedicated signaling, or both.

[0328] Aspect 22: A method for wireless communication at a UE using a first set of parameters associated with a first machine learning model for a task, the method comprising: sending a capability message to a network entity indicating the UE's capability to perform a tuning process of the first machine learning model, wherein the first set of parameters is associated with the first machine learning model; performing the tuning process of the first machine learning model at least in part based on the UE's capability to perform the tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model; sending a message indicating the second set of parameters to the network entity at least in part based on performing the tuning process of the first machine learning model; receiving an allowed status indication from the network entity associated with the second set of parameters; based on the received allowed status indication, performing the task using the first set of parameters associated with the first machine learning model or the second set of parameters associated with the first machine learning model; receiving a not allowed status indication from the network entity associated with the second set of parameters; and based on the received not allowed status indication, performing the task using the first set of parameters associated with the first machine learning model.

[0329] Aspect 23: The method according to aspect 22, the method further comprising: autonomously determining whether to use the first set of parameters associated with the first machine learning model or the second set of parameters associated with the first machine learning model to perform the task at least in part based on the received allowed status indication.

[0330] Aspect 24: A method for wireless communication at a UE using a first set of parameters associated with a first machine learning model for a task, the method comprising: sending a capability message to a network entity indicating the UE's capability to perform an online tuning process of the first machine learning model, wherein the first set of parameters is associated with the first machine learning model; receiving a first indication from the network entity associated with the online tuning process, wherein the first indication includes an activation status or an allowed status; and performing the tuning process of the first machine learning model at least in part based on the UE's capability to perform the online tuning process of the first machine learning model and the received allowed status to obtain a second set of parameters associated with the first machine learning model.

[0331] Aspect 25: A method for wireless communication at a network entity, the method comprising: receiving from a UE a capability message indicating the UE's capability to perform a tuning process of a first machine learning model associated with a first set of parameters at the UE; and receiving from the UE a message indicating at least a part of a second set of parameters.

[0332] Aspect 26: The method according to aspect 25, wherein the first machine learning model includes an encoder part of a second machine learning model, and the third machine learning model includes a decoder part of the second machine learning model.

[0333] Aspect 27: The method according to any one of aspects 25 to 26, wherein receiving the capability message includes: receiving an indication of a set of machine learning models supported by the UE.

[0334] Aspect 28: The method according to any one of aspects 25 to 27, wherein receiving the message is associated with receiving channel state information feedback.

[0335] Aspect 29: The method according to any one of aspects 25 to 28, the method further comprising: sending a first indication associated with performing a tuning process of the first machine learning model to the UE, wherein the first indication includes an activation state or a permission state; and receiving a second indication from the UE in response to the first indication, wherein the second indication includes an activation indication associated with starting to perform the tuning process or a deactivation indication associated with stopping the tuning process.

[0336] Aspect 30: The method according to aspect 29, the method further comprising: receiving an activation request from the UE, wherein sending the first indication is at least partially based on the activation request.

[0337] Aspect 31: The method according to any one of aspects 29 to 30, the method further comprising: sending a third indication to the UE, wherein the third indication includes a deactivation indication associated with stopping the tuning process.

[0338] Aspect 32: An apparatus for wireless communication at a UE, the apparatus comprising: a processor; a memory coupled to the processor; and instructions stored in the memory and executable by the processor to cause the apparatus to perform the method according to any one of aspects 1 to 21.

[0339] Aspect 33: An apparatus for wireless communication at a UE, the apparatus comprising at least one component for performing the method according to any one of aspects 1 to 21.

[0340] Aspect 34: A non-transitory computer-readable medium storing code for wireless communication at a UE, the code including instructions executable by a processor to perform the method according to any one of aspects 1 to 21.

[0341] Aspect 35: An apparatus for wireless communication at a UE using a first set of parameters associated with a first machine learning model for a task, the apparatus comprising: a processor; a memory coupled to the processor; and instructions stored in the memory and executable by the processor to cause the apparatus to perform the method according to any one of aspects 22 to 22.

[0342] Aspect 36: An apparatus for wireless communication at a UE using a first set of parameters associated with a first machine learning model for a task, the apparatus comprising at least one component for performing the method according to any one of aspects 22 to 22.

[0343] Aspect 37: A non-transitory computer-readable medium storing code for wireless communication at a UE using a first set of parameters associated with a first machine learning model for a task, the code comprising instructions executable by a processor to perform the method according to any one of aspects 22 to 22.

[0344] Aspect 38: An apparatus for wireless communication at a UE using a first set of parameters associated with a first machine learning model for a task, the apparatus comprising: a processor; a memory coupled to the processor; and instructions stored in the memory and executable by the processor to cause the apparatus to perform the method according to any one of aspects 24 to 24.

[0345] Aspect 39: An apparatus for wireless communication at a UE using a first set of parameters associated with a first machine learning model for a task, the apparatus comprising at least one component for performing the method according to any one of aspects 24 to 24.

[0346] Aspect 40: A non-transitory computer-readable medium storing code for wireless communication at a UE using a first set of parameters associated with a first machine learning model for a task, the code comprising instructions executable by a processor to perform the method according to any one of aspects 24 to 24.

[0347] Aspect 41: An apparatus for wireless communication at a network entity, the apparatus comprising: a processor; a memory coupled to the processor; and instructions stored in the memory and executable by the processor to cause the apparatus to perform the method according to any one of aspects 25 to 31.

[0348] Aspect 42: An apparatus for wireless communication at a network entity, the apparatus comprising at least one component for performing the method according to any one of Aspects 25 to 31.

[0349] Aspect 43: A non-transitory computer-readable medium storing code for wireless communication at a network entity, the code comprising instructions executable by a processor to perform the method according to any one of Aspects 25 to 31.

[0350] It should be noted that the methods described herein depict possible specific implementations, and the operations and steps may be rearranged or otherwise modified and other specific implementations are possible. Additionally, aspects from two or more methods may be combined.

[0351] Although aspects of LTE, LTE-A, LTE-A Pro, or NR systems may be described for example purposes and the LTE, LTE-A, LTE-A Pro, or NR terminology may be used in much of the description, the techniques described herein are also applicable to networks other than LTE, LTE-A, LTE-A Pro, NR, or advanced 5G networks. For example, the described techniques may be applicable to various other wireless communication systems such as Ultra Mobile Broadband (UMB), Institute of Electrical and Electronics Engineers (IEEE) 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), IEEE 802.20, Flash-OFDM, and other systems and radio technologies not explicitly mentioned herein.

[0352] The information and signals described herein may be represented using any of a variety of different technologies and methods. For example, data, instructions, commands, information, signals, bits, symbols, and chips mentioned throughout the specification may be represented by voltage, current, electromagnetic waves, magnetic fields or magnetic particles, optical fields or optical particles, or any combination thereof.

[0353] The various illustrative blocks and components described in connection with the disclosure herein may be implemented or performed using a general-purpose processor, DSP, ASIC, CPU, FPGA, or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general-purpose processor may be a microprocessor, but in the alternative, the processor may be any processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices (e.g., a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration).

[0354] The functions described herein can be implemented using hardware, software executed by a processor, firmware, or any combination thereof. When implemented in software executed by a processor, the functions can be stored as one or more instructions or code on a computer-readable medium or transmitted using one or more instructions or code on a computer-readable medium. Other examples and specific implementations are within the scope of the present disclosure and the appended claims. For example, due to the nature of software, the functions described herein can be implemented using software executed by a processor, hardware, firmware, hardwiring, or any combination of these items. The features implementing the functions can also be physically located at different positions, including being distributed such that parts of the functions are implemented at different physical positions.

[0355] Computer-readable media includes both non-transitory computer storage media and communication media, which includes any medium that facilitates transfer of a computer program from one location to another. Non-transitory storage media can be any available media that can be accessed by a general or special purpose computer. By way of example, and not limitation, non-transitory computer-readable media can include RAM, ROM, electrically erasable programmable ROM (EEPROM), flash memory, compact disc (CD) ROM or other optical disc storage, magnetic disk storage or other magnetic storage devices, or any other non-transitory media that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a general or special purpose computer or a general or special purpose processor. Additionally, any connection is properly termed a computer-readable medium. For example, if software is transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of computer-readable medium. As used herein, disk and disc include CD, laser disc, optical disc, digital versatile disc (DVD), floppy disk, and Blu-ray disc. Disk can magnetically reproduce data, while disc can optically reproduce data using lasers. Combinations of the above are also included within the scope of computer-readable media.

[0356] As used herein (including in the claims), the "or" used in a listing of items (e.g., a listing of items accompanied by a phrase such as "at least one of" or "one or more of") indicates an inclusive listing such that, for example, the listing of at least one of A, B, or C means A or B or C or AB or AC or BC or ABC (i.e., A and B and C). Additionally, as used herein, the phrase "based on" should not be construed as a reference to a closed set of conditions. For example, an example step described as "based on condition A" can be based on both condition A and condition B without departing from the scope of the present disclosure. In other words, as used herein, the phrase "based on" should be interpreted in the same manner as the phrase "at least partially based on".

[0357] The term "determine" encompasses a variety of actions and, thus, "determine" can include operations, calculations, processing, derivation, investigation, lookup (such as lookup via a table, database, or other data structure), ascertaining, and similar actions. Additionally, "determine" can include receiving (e.g., receiving information), accessing (e.g., accessing data stored in a memory), etc. Additionally, "determine" can include parsing, obtaining, selecting, choosing, establishing, and other such similar actions.

[0358] In the figures, similar components or features may have the same reference numeral. Additionally, various components of the same type can be distinguished by following the reference numeral with a dash and a second numeral used to differentiate among the similar components. If only the first reference numeral is used in the specification, the description can apply to any one of the similar components having the same first reference numeral, regardless of the second reference numeral or any other subsequent reference numerals.

[0359] The description set forth herein in connection with the figures describes example configurations and does not represent all examples that can be implemented or are within the scope of the claims. The term "example" as used herein means "serving as an example, instance, or illustration" and not "preferred" or "advantageous over other examples". The detailed description includes specific details for providing an understanding of the described techniques. However, the techniques can be practiced without these specific details. In some instances, well-known structures and devices are shown in block diagram form to avoid obscuring the concepts of the described examples.

[0360] The present description is provided to enable a person of ordinary skill in the art to make or use the present disclosure. Various modifications to the present disclosure will be apparent to a person of ordinary skill in the art, and the general principles defined herein can be applied to other variations without departing from the scope of the present disclosure. Thus, the present disclosure is not limited to the examples and designs described herein, but should be accorded the broadest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for wireless communication at a user equipment (UE), the method comprising: Obtaining a data sample associated with a task at the UE, wherein a first set of parameters is associated with the first machine learning model; Sending a capability message indicating the UE's ability to perform a tuning process of the first machine learning model; Performing the tuning process of the first machine learning model at least in part based on the UE's ability to perform the tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model; And Sending a message indicating at least a portion of the second set of parameters to a network entity at least in part based on performing the tuning process of the first machine learning model.

2. The method according to claim 1, wherein the first machine learning model comprises an encoder portion of a second machine learning model, and a third machine learning model comprises a decoder portion of the second machine learning model.

3. The method according to claim 2, wherein performing the tuning process of the first machine learning model further comprises: Receiving a set of parameters associated with a loss function.

4. The method according to claim 3, the method further comprising: Sending a message associated with a forward propagation process.

5. The method according to claim 3, the method further comprising: Receiving a message associated with a backpropagation process for adjusting parameters associated with an encoder, wherein the message indicates a gradient associated with the loss function; And Updating the parameters associated with the encoder at least in part based on the message.

6. The method according to claim 3, the method further comprising: Receiving the set of parameters associated with the loss function; And Updating parameters associated with the decoder portion of the second machine learning model at least in part based on the set of parameters.

7. The method according to claim 1, wherein sending the capability message comprises: Sending an indication of a set of machine learning models supported by the UE.

8. The method according to claim 1, wherein the task comprises a channel state information (CSI) feedback task.

9. The method according to claim 1, wherein performing the tuning process of the first machine learning model further comprises: Using the second set of parameters to update parameters associated with an encoder, parameters associated with a decoder, or both.

10. The method according to claim 9, wherein performing the tuning process of the first machine learning model comprises: Performing a tuning process.

11. The method according to claim 10, wherein performing the tuning process of the first machine learning model comprises: When performing the task, using the second set of parameters for the first machine learning model to update the second set of parameters associated with the encoder.

12. The method according to claim 9, wherein performing the tuning process of the first machine learning model comprises: Performing an offline tuning process.

13. The method according to claim 12, wherein performing the offline tuning process of the first machine learning model comprises: When performing the task, updating the second parameter set associated with the encoder using the first parameter set for the first machine learning model.

14. The method according to claim 12, the method further comprising: Sending a third indication to the network entity, wherein the third indication indicates the availability of a second encoder, and wherein the second encoder is associated with performing the offline tuning process.

15. The method according to claim 1, wherein performing the tuning process further comprises: Receiving a first indication from the network entity associated with performing the tuning process of the first machine learning model, wherein the first indication comprises an activation state or an allowed state; And Sending a second indication to the network entity in response to the first indication, wherein the second indication comprises an activation indication associated with starting to perform the tuning process or a deactivation indication associated with stopping the tuning process.

16. The method according to claim 15, the method further comprising: Sending an activation request to the network entity, wherein receiving the first indication is at least partially based on the activation request.

17. The method according to claim 15, the method further comprising: Receiving a fourth indication from the network entity, wherein the fourth indication comprises a deactivation indication associated with stopping the tuning process.

18. The method according to claim 1, wherein performing the tuning process of the first machine learning model further comprises: Updating the parameters associated with the encoder, the parameters associated with the decoder, or both using the second parameter set at least partially based on the gradients associated with the first parameter set and the second parameter set.

19. The method according to claim 18, wherein updating the parameters associated with the encoder, the parameters associated with the decoder, or both further comprises: Receiving a parameter set associated with a loss function; And Sending a message associated with the forward propagation process.

20. The method according to claim 19, the method further comprising: Receiving a message associated with a backpropagation process for adjusting the parameters associated with the encoder, the parameters associated with the decoder, or both, wherein the message indicates the gradient associated with the loss function; and Updating the parameters associated with the encoder, the parameters associated with the decoder, or both at least partially based on the message.

21. The method according to claim 1, wherein the UE receives the indication of the first parameter set via broadcast signaling, dedicated signaling, or both.

22. A method for wireless communication at a user equipment (UE) using a first parameter set associated with a first machine learning model for a task, the method comprising: Sending a capability message indicating the UE's capability to perform a tuning process of the first machine learning model to a network entity, wherein the first parameter set is associated with the first machine learning model; Performing the tuning process of the first machine learning model at least partially based on the ability of the UE to perform the tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model; Sending a message indicating the second set of parameters to a network entity at least partially based on performing the tuning process of the first machine learning model; Receiving an allowed status indication from the network entity associated with the second set of parameters; Performing the task using the first set of parameters associated with the first machine learning model or the second set of parameters associated with the first machine learning model at least partially based on the received allowed status indication; Receiving a not-allowed status indication from the network entity associated with the second set of parameters; And Performing the task using the first set of parameters associated with the first machine learning model based on the received not-allowed status indication.

23. The method according to claim 22, the method further comprising: Autonomously determining whether to use the first set of parameters associated with the first machine learning model or the second set of parameters associated with the first machine learning model to perform the task at least partially based on the received allowed status indication.

24. A method for wireless communication at a user equipment (UE) using a first set of parameters associated with a first machine learning model for a task, the method comprising: Sending an ability message indicating the ability of the UE to perform a tuning process of the first machine learning model, wherein the first set of parameters is associated with the first machine learning model; Receiving a first indication from the network entity associated with the tuning process, wherein the first indication includes an activation status or an allowed status; and Performing the tuning process of the first machine learning model to obtain a second set of parameters associated with the first machine learning model at least partially based on the ability of the UE to perform the tuning process of the first machine learning model and the received allowed status.

25. A method for wireless communication at a network entity, the method comprising: Receiving an ability message from a user equipment (UE) indicating the ability of the UE to perform a tuning process of a first machine learning model associated with a first set of parameters at the UE; And Receiving a message indicating at least a portion of a second set of parameters from the UE.

26. The method according to claim 25, wherein the first machine learning model includes an encoder portion of a second machine learning model, and a third machine learning model includes a decoder portion of the second machine learning model.

27. The method according to claim 25, wherein receiving the ability message comprises: Receiving an indication of a set of machine learning models supported by the UE.

28. The method according to claim 25, wherein receiving the message is associated with receiving channel state information feedback.

29. The method according to claim 25, the method further comprising: Send a first indication associated with performing the tuning process of the first machine learning model to the UE, where the first indication includes an active state or an allowed state; And Receive a second indication from the UE in response to the first indication, where the second indication includes an activation indication associated with starting to perform the tuning process or a deactivation indication associated with stopping the tuning process.

30. The method according to claim 29, the method further comprising: Receive an activation request from the UE, where sending the first indication is at least partially based on the activation request.

Images