Method for reducing diversity of bilateral machine learning model

Abstract

There is provided an apparatus comprising means for performing: transmitting, to a network node, information about at least one two-sided machine learning model supported by the apparatus; or transmitting a machine learning profile identification of the device to a network node, wherein the machine learning profile identification is associated with at least one two-sided machine learning model supported by the device; wherein the at least one bilateral machine learning model is configured to enable the device and the network node to perform joint reasoning; an indication of at least one first selected two-sided machine learning model supported by the apparatus is received from a network node.

Description

Technical Field

[0001] Several example embodiments are directed to a method of reducing the diversity of two-sided machine learning models inferred by user devices and network nodes. Background Art

[0002] Machine learning (ML) techniques can be used for various tasks in communication networks. A two-sided ML model is a model on which joint reasoning can be performed between user devices and network nodes. Summary of the invention

[0003] According to some aspects, the subject matter of independent claims is provided. Some example embodiments are defined in the dependent claims. The scope of protection of the various example embodiments is defined by the independent claims. The example embodiments and features described in this specification that do not fall within the scope of the independent claims should be interpreted as examples that help understand the various example embodiments.

[0004] According to a fourth aspect, a method is provided, comprising: transmitting, by a device, information about at least one two-sided machine learning model supported by the device to a network node; or transmitting, by a device, a machine learning profile identifier of the device to a network node, wherein the machine learning profile identifier is associated with at least one two-sided machine learning model supported by the device; wherein the at least one two-sided machine learning model is configured to enable the device and the network node to perform joint reasoning; and receiving, by the device, an indication of at least one first selected two-sided machine learning model supported by the device from the network node.

[0005] According to one embodiment, the method includes starting joint inference with a network node using a first selected two-sided machine learning model.

[0006] According to one embodiment, information about at least one two-sided machine learning model includes: an identification of at least one two-sided machine learning model supported by the device; and information about functionality of at least one two-sided machine learning model supported by the device.

[0007] According to one embodiment, the method includes receiving an instruction from a network node to switch to a second two-sided machine learning model supported by the device, wherein the second two-sided machine learning model is also supported by another device served by the network node.

[0008] According to one embodiment, the method includes: switching to a second two-sided machine learning model; and starting joint reasoning with the network nodes using the second two-sided machine learning model.

[0009] According to one embodiment, the method includes: determining that the device no longer supports the second two-sided machine learning model; based on the determination, transmitting a negative response to the network node; and continuing joint reasoning with the first selected two-sided machine learning model.

[0010] According to one embodiment, the method includes: establishing an auxiliary connection with another network node; transmitting information about at least one two-sided machine learning model supported by the device to the other network node; or transmitting a machine learning profile identifier of the device to the other network node, wherein the machine learning profile identifier is associated with at least one two-sided machine learning model supported by the device; and starting joint reasoning with the other network node using a first selected two-sided machine learning model.

[0011] According to one embodiment, the method includes: receiving an instruction from a network node to switch to a second two-sided machine learning model supported by the device, and the second two-sided machine learning model is also supported by the other network node; switching to the second two-sided machine learning model; and starting joint reasoning with the network node and the other network node using the second two-sided machine learning model.

[0012] According to one embodiment, the method includes: receiving an indication of a second selected two-sided machine learning model from the other network when performing joint inference with the network node using the first selected two-sided machine learning model; transmitting a negative response to the other network, and an indication of the device performing joint inference with the network node using the first selected two-sided machine learning model.

[0013] According to one embodiment, the method includes: receiving an instruction from the other network node to start joint inference using the first selected two-sided machine learning model, or an instruction to refuse to perform joint inference using the first selected two-sided machine learning model.

[0014] According to a fifth aspect, a method is provided, comprising: receiving, by a network node, information about at least one bilateral machine learning model supported by the user device from a user device; or receiving, by a network node, a machine learning profile identifier of the user device from the user device, wherein the machine learning profile identifier is associated with at least one bilateral machine learning model supported by the user device; and obtaining, by the network node, information about at least one bilateral machine learning model supported by the user device based on the machine learning profile identifier of the user device; wherein the at least one bilateral machine learning model is configured to enable the apparatus and the user device to perform joint reasoning; selecting, by the network node, a first bilateral machine learning model supported by the user device; and transmitting, by the network node, an indication of selecting at least the first bilateral machine learning model supported by the user device to the user device.

[0015] According to one embodiment, the method includes: starting joint inference with a user device using a first two-sided machine learning model.

[0016] According to one embodiment, the first two-sided machine learning model is selected based on other two-sided machine learning models that are currently being jointly inferred by the apparatus and one or more other user devices, or other two-sided machine learning models supported by other user devices.

[0017] According to one embodiment, the method includes: receiving information about at least one bilateral machine learning model supported by another user device from another user device served by the apparatus; or receiving a machine learning profile identifier of another user device from another user device served by the apparatus, wherein the machine learning profile identifier is associated with at least one bilateral machine learning model supported by the other user device; and obtaining information about at least one bilateral machine learning model supported by the other user device based on the machine learning profile identifier of the other user device.

[0018] According to one embodiment, the method includes: determining a second bilateral machine learning model supported by both user devices based on information about at least one bilateral machine learning model supported by the user device and the other user device; and transmitting an instruction to the user device to switch to the second bilateral machine learning model.

[0019] According to one embodiment, the method includes: transmitting an indication of selecting a second two-sided machine learning model to the other user device; and starting joint reasoning with the user device and the other user device using the second two-sided machine learning model.

[0020] According to one embodiment, the method includes: receiving a negative response from the user device, indicating that the user device no longer supports the second two-sided machine learning model; continuing reasoning with the user device using the first two-sided machine learning model; and starting joint reasoning with the other user device using the second two-sided machine learning model.

[0021] According to one embodiment, the method includes: receiving a request for information about a two-sided machine learning model of the user equipment from another network node serving the user equipment; determining that a first two-sided machine learning model is compatible with capabilities of the other network node; and transmitting an indication of selecting the first two-sided machine learning model to the other network node.

[0022] According to one embodiment, the method includes: receiving a request for information about a two-sided machine learning model of the user device from another network node serving the user device; determining that a first two-sided machine learning model is incompatible with capabilities of the other network node; based on the determination, selecting a second two-sided machine learning model that is compatible with the capabilities of the user device and the capabilities of the other network node; and transmitting an instruction to the user device to switch to the second two-sided machine learning model.

[0023] According to one embodiment, the method comprises: transmitting an indication of selecting a second two-sided machine learning model to the other network node; and starting inference with the user equipment and the other network node using the second two-sided machine learning model.

[0024] According to a sixth aspect, a method is provided, comprising: establishing a connection with a user equipment, the device being configured to serve it as an auxiliary network node; receiving information about a two-sided machine learning model that is compatible with the capabilities of the user equipment and the capabilities of the auxiliary network node from a main network node configured to serve the user equipment; and starting joint reasoning with the user equipment and the main network node using the two-sided machine learning model.

[0025] According to one aspect, there is provided an apparatus comprising means for executing the method of the fourth aspect and any embodiments thereof.

[0026] According to one aspect, a device is provided, comprising means for executing the method of the fifth aspect and any embodiment thereof.

[0027] According to one aspect, there is provided an apparatus comprising means for executing the method of the sixth aspect and any embodiment thereof.

[0028] According to one embodiment, the device (in any of the above aspects) includes at least one processor; and at least one memory can store instructions, which, when executed by the at least one processor, cause the device to perform corresponding operations.

[0029] According to one aspect, a non-transitory computer-readable medium containing instructions is provided. When the instructions are executed by an apparatus, the apparatus is caused to perform the method of at least the fourth aspect and any embodiments thereof.

[0030] According to one aspect, a non-transitory computer-readable medium containing instructions is provided. When the instructions are executed by an apparatus, the apparatus performs the method of at least the fifth aspect and any embodiment thereof.

[0031] According to one aspect, a non-transitory computer-readable medium containing instructions is provided. When the instructions are executed by an apparatus, the apparatus performs the method of at least the sixth aspect and any embodiment thereof.

[0032] According to one aspect, a computer program comprising instructions is provided. When the instructions are executed by an apparatus, the apparatus is caused to perform the method of the fourth aspect and any embodiment thereof.

[0033] According to one aspect, a computer program comprising instructions is provided. When the instructions are executed by an apparatus, the apparatus performs the method of the fifth aspect and any embodiment thereof.

[0034] According to one aspect, a computer program comprising instructions is provided, which, when executed by an apparatus, causes the apparatus to perform the method of the sixth aspect and any embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Some example embodiments will now be described with reference to the accompanying drawings.

[0036] Figure 1 An example of a network architecture for a communication system is shown;

[0037] Figure 2a An example of a network node serving multiple user equipments is shown;

[0038] Figure 2b An example of a user device in a multi-connection mode is shown;

[0039] Figure 3 An example of a method flow chart is shown;

[0040] Figure 4 An example of a method flow chart is shown;

[0041] Figure 5 An example of signaling between entities is shown;

[0042] Figure 6 An example of signaling between entities is shown;

[0043] Figure 7a An example of signaling between entities is shown;

[0044] Figure 7b An example of signaling between entities is shown;

[0045] Figure 8 An example of signaling between entities is shown;

[0046] Fig. 9 An example of signaling between entities is shown;

[0047] Fig.10 An example of signaling between entities is shown;

[0048] Fig.11 shows an example of signaling between entities; and

[0049] Fig.12 An example of a block diagram of an apparatus is shown. DETAILED DESCRIPTION

[0050] Figure 1 An example of a network architecture of a communication system is shown. Different exemplary embodiments will be described below using LTE Advanced (LTE-A) or New Radio (NR, also known as fifth generation (5G)) as an example of a wireless access architecture, but the embodiments are not limited to such architectures. It is obvious to those skilled in the art that these embodiments may also be applied to other types of communication networks with suitable devices by appropriately adjusting parameters and processes. Examples of other suitable systems include Universal Mobile Telecommunications System (UMTS) Radio Access Network (UTRAN or E-UTRAN), Long Term Evolution (LTE, the same as E-UTRA), Wireless Local Area Network (WLAN or WiFi), Worldwide Interoperability for Microwave Access (WiMAX), Personal Communications Service (PCS), Wideband Code Division Multiple Access (WCDMA), systems using Ultra-Wideband (UWB) technology, sensor networks, Mobile Ad Hoc Networks (MANETs), and Internet Protocol Multimedia Subsystem (IMS), or any combination thereof.

[0051] Figure 1 The example of FIG. 1 shows a portion of an exemplary wireless access network. Figure 1User equipment (UE) 100 and 102 are shown, which are configured to be wirelessly connected to an access node 104 (such as a gNB, i.e., next generation NodeB, or an eNB, i.e., evolved NodeB (eNodeB)) providing the cell on one or more communication channels in a cell. The physical link from the user equipment to the network node is called the uplink (UL) or reverse link, and the physical link from the network node to the user equipment is called the downlink (DL) or forward link. It should be understood that the network node or its functions can be implemented by any suitable node, host, server or access point. The communication system usually includes multiple network nodes, in which case the network nodes can also be configured to communicate with each other through wired or wireless links designed for this purpose. These links can be used for signaling purposes. A network node is a computing device that is configured to control the radio resources of the communication system to which it is coupled. A network node may also be referred to as a base station (BS), access point or any other type of An interface device of a type, which includes a relay station capable of operating in a wireless environment. The network node includes or is coupled to a transceiver. A connection to an antenna unit is provided from the transceiver of the network node, and the antenna unit establishes a bidirectional radio link to a user equipment. The antenna unit may include multiple antennas or antenna elements. The network node is further connected to a core network 110 (CN) or a next generation core network (NGC). Depending on the system, the corresponding device on the CN side may be a serving gateway (S-GW, used to route and forward user data packets), a packet data network gateway (P-GW), used to provide a user equipment (UE) with a connection to an external packet data network, or a mobility management entity (MME), etc. An example of a network node configured to operate as a relay station is an integrated access and backhaul node (IAB). The distributed unit (DU) part of the IAB node performs the BS function of the IAB node, while the backhaul connection is performed by the mobile terminal (MT) part of the IAB node. The UE function can be performed by the IAB The MT performs the BS function, and the BS function can be performed by the IAB DU. The network architecture can include a parent node, namely the IAB donor, which can be connected to the CN by wire and to the IAB MT by wireless.

[0052] User equipment, or user terminal (UE) generally refers to a portable computing device, including wireless mobile communication devices operating with or without a subscriber identity module (SIM), including but not limited to the following types of devices: mobile stations (mobile phones), smart phones, personal digital assistants (PDAs), handheld devices, devices using wireless modems (alarm or measurement equipment, etc.), portable computers and / or touch screen computers, tablets, game consoles, laptops and multimedia devices. It should be understood that a user equipment can also be a device that is almost only used for uplink, such as a camera or video camera that uploads images or video clips to a network. A user equipment can also be a device with the ability to operate in the Internet of Things (IoT), which is a scenario in which objects can transmit data over a network without the need for human-to-human or human-to-computer interaction.

[0053] Furthermore, although the devices are described as a single entity, different units, processors and / or storage units may be arranged within these devices. Figure 1 (not all shown) to achieve its function.

[0054] 5G enables the use of multiple-input multiple-output (MIMO) technology on both the UE and gNB sides, with a larger number of base stations or nodes than LTE (the so-called small cell concept), including macro sites operating in collaboration with smaller sites, and the use of various radio technologies depending on service requirements, use cases and / or available spectrum. 5G mobile communications support a wide range of use cases and related applications, including video streaming, augmented reality, different ways of sharing data, and various forms of machine-type applications (such as (massive) machine-type communications (mMTC)), including vehicle safety, different sensors and real-time control. 5G is expected to have multiple radio interfaces, namely below 7 GHz, centimeter wave and millimeter wave, and also be able to integrate with existing legacy radio access technologies (such as LTE). The frequency range below 7 GHz can be referred to as FR1, and above 24 GHz (or more precisely 24-52.6 GHz) can be referred to as FR2. Integration with LTE can be realized, at least in the early stages, as a system where LTE provides macro coverage and 5G radio interface access is achieved by small cells aggregated to LTE. In other words, 5G plans to support inter-radio access technology (Inter-RAT) interoperability (such as LTE-5G) and inter-radio interface (Inter-RI) interoperability (such as sub-7GHz-cmWave, sub-7GHz-cmWave-mmWave). One of the concepts considered for use in 5G networks is network slicing, in which multiple independent and dedicated virtual subnets (network instances) can be created within the same infrastructure to run services with different requirements for latency, reliability, throughput, and mobility.

[0055] The communication system can also communicate with other networks (such as the public switched telephone network or the Internet 112) or utilize services provided by them. The communication network can also support the use of cloud services, for example, at least part of the core network operation can be performed as a cloud service (this is in Figure 1 The communication system may also include a central control entity or similar entity providing facilities for the networks of different operators to cooperate, for example in spectrum sharing.

[0056] By leveraging network function virtualization (NVF) and software defined networking (SDN), edge cloud can be introduced into the radio access network (RAN). Using edge cloud may mean that the access node performs operations at least partially in a server, host or node, which is operatively coupled to a remote radio head or a base station including radio components. The operation of the node can also be distributed among multiple servers, nodes or hosts. The application of cloud RAN architecture enables RAN real-time functions to be performed on the RAN side (in the distributed unit DU 104) and non-real-time functions to be performed in a centralized manner (in the centralized unit CU108).

[0057] 5G can also make use of satellite communications to enhance or supplement the coverage of 5G services, for example by providing backhaul transport. Possible use cases include providing service continuity for machine-to-machine (M2M) or Internet of Things (IoT) devices or passengers on a vehicle, or ensuring service availability for critical communications and future rail / maritime / aeronautical communications. Satellite communications can make use of geostationary orbit (GEO) satellite systems as well as low earth orbit (LEO) satellite systems, in particular mega constellations (systems with hundreds of (nano) satellites deployed). Each satellite 106 in the constellation can cover multiple satellite communication-capable network entities, which create ground cells. Ground cells can be created by ground relay nodes 104 or gNBs located on the ground or on satellites.

[0058] Radio Access Network (RAN) optimization algorithms may include algorithms for optimizing and / or improving the operation, performance and / or one or more functions of the RAN. RAN optimization may include, for example, increasing or decreasing the priority of a service. RAN optimization for end-user perceived improvement includes, for example, capacity and coverage optimization, load sharing, load balancing, random access channel (RACH) optimization, and energy saving. These functions may be optimized by self-organizing network (SON) algorithms. SON refers to the ability of a network to work in a self-organizing manner. Modern RAN features are expected to have capabilities such as: self-planning, self-configuration, and self-optimization.

[0059] The radio access network optimization algorithm can be implemented by, for example, machine learning techniques. The application of machine learning (ML) in SON solutions is also feasible.

[0060] Machine learning (ML) refers to algorithms and statistical models used by computer systems to perform specific tasks, which do not rely on explicit instructions, but rather on models and reasoning. It is considered a subset of artificial intelligence (AI). Machine learning algorithms build mathematical models based on sample data (called "training data") to make predictions or decisions without being explicitly programmed to perform tasks. ML algorithms can be divided into, for example, supervised learning, unsupervised learning, and reinforcement learning. An ML algorithm may consist of one or more ML components, forming a so-called ML workflow (pipeline), where each component can be placed in a different RAN network function and / or the UE itself for execution.

[0061] A two-sided ML model refers to a pair of AI / ML models on which joint reasoning can be performed between the UE and the network. For example, the first part of the reasoning can be performed by the UE first, and the rest by the network node (such as the gNB), or vice versa.

[0062] An example of a two-sided ML model is channel state information (CSI) feedback compression using an autoencoder. The encoder can be implemented by the UE and the decoder can be implemented by a network node (e.g., a gNB). The encoder receives the CSI H as input. The total number of feedback parameters in H is Nt×Nr×Nc, where Nt is the number of transmit antennas at the gNB, Nr is the number of receive antennas at the UE, and Nc is the number of subcarriers on which the orthogonal frequency division multiplexing (OFDM) system operates. The neural network in the encoder part compresses H into a codeword S of smaller size. In one example, S can be 64 bits. The neural network in the decoder part receives S as input and recovers the H matrix.

[0063] The autoencoder can be trained as a whole, or the encoder and decoder parts can be trained separately. In the case of separate training, the encoder part can be trained using, for example, an unsupervised algorithm. Once the encoder is trained, it can be supervised to train the decoder using the generated labeled data.

[0064] Autoencoders using bilateral ML models for CSI feedback compression have been shown to improve compression speed.

[0065] Figure 2aAn example of a network node serving multiple user equipments (e.g., UE 1 220 and UE X 225) is shown. The network node 210 (e.g., a gNB) can perform full or joint reasoning for various UE functions using ML models located at the network node 210 or split between the UEs 220, 225 and the network node 210. Due to the large number of UEs and the diverse functions, the number of ML models running on the network node 210 can be very large. Each ML model 1...X consumes resources 230 on the network node 210, including at least random access memory (RAM) and the capacity of the ML accelerator. The large consumption of resources by these ML models can result in significant and unpredictable hardware resource requirements and potentially lead to shortages of these resources.

[0066] Figure 2b An example of a user equipment in a multi-connection mode is shown. A UE 240 may be in a multi-connection mode with multiple access points (APs) (e.g., AP1 250 and APX 255) or network nodes 250, 255. The UE 240 may perform full or joint reasoning for various UE functions using ML models located in the UE 240 or separated between the UE 240 and the access points 250, 255. Running different ML models for the same function for different connections consumes a significant amount of resources 260 of the UE 240.

[0067] Methods are proposed to reduce the diversity of ML models running simultaneously on UEs and / or network nodes without affecting functional performance.

[0068] Figure 3 An example of a flow chart of a method 300 is shown. The various stages of the method 300 shown may be performed by a UE, or by a control device configured to control its functions (when installed therein). The UE may be, for example, Figure 5 The device 510, or Figure 2bUE 240 in, which is configured to perform at least method 300. The method includes 310a: transmitting, by the device, information about at least one two-sided machine learning model supported by the device to a network node. Alternatively, the method includes 310b: transmitting, by the device, a machine learning profile identifier (or similarly a machine learning profile identifier) ​​of the device to the network node, wherein the machine learning profile identifier is associated with at least one two-sided machine learning model supported by the device. In general, a machine learning profile identifier (or identifier) ​​is an identifier or identifier used to obtain information about the machine learning capabilities of the device (i.e., about the machine learning profile of the device), including a list of one or more two-sided machine learning models supported by the device. In other words, the machine learning profile identifier (or identifier) ​​is used to identify the machine learning profile of the device (i.e., the machine learning capabilities of the device). The at least one two-sided machine learning model is configured to enable the device and the network node to perform joint reasoning. Method 300 includes 320: receiving, by the device from the network node, an indication related to at least one first selected two-sided machine learning model supported by the device.

[0069] Figure 4 An example of a flow chart of a method 400 is shown. The various stages of the method 400 shown may be performed by a network node, or by a control device configured to control its functions (when installed therein). The UE may be, for example, Figure 5 Device 520, or Figure 2b A network node 210 (e.g., a gNB) in the apparatus is configured to perform at least method 400. Method 400 includes 410a: receiving, by the network node, information about at least one bilateral machine learning model supported by the user equipment from a user equipment. Alternatively, method 400 includes 410b: receiving, by the network node, a machine learning profile identifier of the user equipment from the user equipment, wherein the machine learning profile identifier is associated with at least one bilateral machine learning model supported by the user equipment; and obtaining, by the network node, information about at least one bilateral machine learning model supported by the user equipment based on the machine learning profile identifier of the user equipment. The at least one bilateral machine learning model is configured to enable the apparatus and the user equipment to perform joint reasoning. The method includes 420: selecting, by the network node, a first bilateral machine learning model supported by the user equipment. The method includes 430: transmitting, by the network node to the user equipment, an indication of selecting at least a first bilateral machine learning model supported by the user equipment.

[0070] The method disclosed herein is able to reduce the diversity of two-sided ML models.

[0071] The following will be explained in the context of the signaling diagram Figure 3 and Figure 4 Method and embodiments thereof.

[0072] Figure 5 An example of signaling between entities is shown. The UE 510 transmits 530 to the network node 520 information about at least one bilateral ML model supported by the UE 510. In other words, the UE 510 provides the network node 520 with information about its ML capabilities. The UE may have one or more ML-supported functions, such as optimal beam prediction in the spatial domain and CSI compression. These functions (e.g., each function) may be associated with one or more ML models. The selection of ML models for the functions may depend on requirements or input conditions. These requirements may include, for example, the size of the channel feedback after encoder compression, the computational complexity of the ML model, and the size of the ML model. The complexity of the ML model may be measured in floating point operations (FLOPs). The size of the ML model may be given in megabytes (Mbyte).

[0073] For example, information about ML models may be included in a list. The list may include functions to which different ML models are applicable. For example, the list may include a function identifier (ID) and an ID of an ML model that supports the function. For example, the list may include a [function ID; supported ML model ID] pair.

[0074] As an alternative to providing information about supported ML models, the UE may provide its ML profile identification, such as Figure 6 as shown in the situation.

[0075] The network node 520 stores the information about the supported ML models in a memory. The received information may be paired with the UE ID of the UE 510. The UE ID includes, for example, information about the UE manufacturer or vendor, UE model, and network identification.

[0076] The network node 520 may receive information about ML models from a plurality of UEs connected to the network node 520. The network node 520 may maintain a database of ML models for different functions supported by different UEs.

[0077] The network node 520 selects 532 an ML model for reasoning with the UE 510. The selected ML model may be a first two-sided ML model. The selection may be based on other UE-specific models currently being reasoned with by the network node 520. For example, the network node 520 may select an ML model currently being used with one or more other UEs for the same functionality. This enables a reduction in the diversity of two-sided ML models used for simultaneous reasoning by different UEs. Selecting and using an ML model for the UE 510 that is already used or supported by other UEs served by the network node 520 may reduce the diversity of ML models running simultaneously at the network node.

[0078] The network node 520 transmits 534 to the UE 510 at least the selected ML model, or at least an indication of a first two-sided ML model supported by the UE 510 .

[0079] The UE 510 receives 534 an indication of the selected ML model. The UE 510 starts inference 536 with the network node 520 using the first two-sided ML model.

[0080] Figure 6 An example of signaling between entities is shown. UE1 610 transmits 630 an ML profile ID of UE 610 to network node 620. The ML profile ID is associated with at least one bilateral ML model supported by UE1 610. The ML profile ID may be, for example, in the form of a UE identifier, such as an International Mobile Equipment Identity (IMEI). As another example, the ML profile ID may be a dedicated ID for indicating the ML models supported by the UE.

[0081] After receiving 630 the ML profile ID of UE1 610, the network node 620 can retrieve or obtain information about at least one two-sided ML model supported by UE1 610. The network node 620 may retrieve 631 information about ML models supported by UE1 610 based on the ML profile ID. The information may be included in the ML profile of UE1.

[0082] The network node 620 may obtain the ML profile from a server, such as an operation, administration, and maintenance (O&M) server 622. The O&M server 622 may receive a request for an ML profile corresponding to the ML profile ID of the UE1 610 from the network node 620. The O&M server 622 may obtain the requested information about the ML capability of the UE1 from the vendor's server 622.

[0083] The network node 620 receives information about ML models supported by UE1 from a server (eg, from the O&M server 622 ).

[0084] The network node 620 stores information about the ML models supported by UE1 610 in a memory.

[0085] The network node 620 selects 632 an ML model for inference with UE1 610. The selected ML model may be a first two-sided ML model. The selection may be based on other UE-specific models currently being inferred at the network node 620. For example, the network node 620 may select an ML model currently being used with one or more other UEs for the same functionality. This enables a reduction in the diversity of two-sided ML models being inferred for multiple UEs simultaneously. Selecting to use an ML model for UE 1 610 that is already used or supported by other UEs served by the network node 620 can reduce the diversity of ML models running simultaneously on the network node.

[0086] The network node 620 transmits 634 to the UE1 610 at least the selected ML model or at least an indication of a first two-sided ML model supported by the UE 610 .

[0087] UE1 610 receives 634 an indication of the selected ML model. UE1 610 begins reasoning 636 with the network node 620 using the first two-sided ML model.

[0088] Figure 7a An example of signaling between entities is shown. Figure 7a In the example of , multiple UEs have been connected to the network node 720. For example, UE1 710 and UE2 722 have joined the same network node 720. Figure 7a In the example of , it is assumed that UE1 and UE2 support at least one ML model supported by both UE1 and UE2. Such an ML model supported by both UE1 and UE2 may be referred to as a common ML model.

[0089] The network node 720 receives 730 information about ML models supported by UE1 710. For example, UE1 710 supports ML models 1 and 2.

[0090] The network node 720 stores the received information and performs ML model selection 732. Figure 7a In the example, ML model 1 is selected.

[0091] The network node 720 transmits 734 to the UE1 710 an indication of the selected ML model.

[0092] UE1 710 and network node 720 start reasoning 736 with ML model 1.

[0093] The network node 720 receives 738 information about ML models supported by another user equipment UE2 722. For example, UE2 722 supports ML models 2 and 3.

[0094] Alternatively, the network node 720 may receive the ML profile ID of UE2 722, based on which the network node 720 may obtain or acquire information about supported ML models from the server, such as Figure 6 as shown in the situation.

[0095] The network node 720 performs ML model selection based on information about ML models supported by UE1 710 and ML models supported by UE2 722. Figure 7a In the example of , UE1 710 and network node 720 run two-sided ML model 1 (first two-sided ML model). However, UE2 722 does not support ML model 1.

[0096] The network node 720 determines a second two-sided ML model supported by both user equipments (ie, UE1 and UE2) based on information about the two-sided ML models supported by UE1 710 and UE2 722. The network node 720 determines a common ML model that can be used for reasoning between the two UEs. Figure 7a In the example of , the common model is ML model 2 (the second two-sided ML model).

[0097] The network node 720 transmits 742 to UE1 710 an instruction to switch to ML model 2. UE1 710 receives 742 an instruction to switch to another ML model supported by UE1, which is also supported by another UE served by the network node 720. UE1 710 switches to ML model 2 (a second dual-sided ML model).

[0098] The network node 720 transmits 744 an indication of the selected ML model (ie, ML Model 2) to UE2 722 .

[0099] The network node 720 starts 746 inference with ML model 2 with UE1 710 and UE2 722. Selecting a common ML model for two or more UEs, the network node 720 reduces the resources used for ML model inference. Coordinating inference of ML models for different UEs enables the network node 720 to save resources.

[0100] It is assumed that the network node 720 maintains up-to-date information about the ML models supported by the UEs connected to it. The UE may inform the network node of any changes in the information about the ML models it supports. Figure 7b An example is shown in which the network node 720 holds an outdated ML model list for a certain UE.

[0101] Figure 7bAn example of signaling between entities is shown. Figure 7b The signaling then Figure 7a Signaling. Figure 7b In the example of , UE1 receives 742 an instruction to switch to ML Model 2, but it determines that UE1 710 no longer supports ML Model 2. Based on the determination that UE1 710 does not support ML Model 2, UE1 transmits a negative acknowledgement (NACK) to the network node 720.

[0102] UE1 710 continues 749 reasoning with the network node 720 using the first selected two-sided ML model.

[0103] The network node 720 receives 743 a NACK from UE1 710. The NACK indicates to the network node 720 that UE1 710 no longer supports ML Model 2. The network node 720 may update the database accordingly.

[0104] The network node 720 may decide to run different ML models for UE1 and UE2. Following the indication 744, the network node 720 may continue reasoning with UE1 749 using ML model 1 and start reasoning 747 with UE2 722 using ML model 2. To this end, the network node 720 may want to ensure that it has sufficient resources to run different ML models for different UEs. As an alternative to running different ML models for different UEs, the network node 720 may decide to switch one or more UEs to a non-ML algorithm.

[0105] The UE can support multiple simultaneous connections with the network. Figure 2b As shown. For example, a radio access network (RAN) may support NR-NR dual connectivity (NR-DC), where a UE may be connected to one network node (e.g., a gNB) as a master node (MN) and another network node (e.g., a gNB) as a secondary node (SN). The master node may be a first network node and the secondary node may be a second network node.

[0106] As another example, NR-DC may also be used when the UE is connected to a single network node that acts as both a MN and a SN and configures both a Master Cell Group (MCG) and a Secondary Cell Group (SCG).

[0107] Figure 8 An example of signaling between entities is shown. UE 810 is performing reasoning 830 with ML model 1 with network node 1 820, which may be a first network node or a master node (MN).

[0108] The UE 810 establishes 832 an auxiliary connection with a network node 2 822, which may be a second network node or an auxiliary node (SN) or another network node. The UE 810 transmits information about at least one two-sided ML model supported by the UE 810 to the network node 2 822. The network node 2 822 may maintain a list of ML models supported by the UEs connected to it.

[0109] Network node 2 822 requests 834 from network node 1 information about the ML model for UE 810. For example, network node 2 822 may request an ML model ID for UE 810, such as the ML model ID currently used for UE 810.

[0110] If UE 810 has already started inference with network node 1 820 before establishing the auxiliary connection with network node 2 822, then when indicating its ML capabilities at step 832, the UE may indicate to network node 2 822 the ML model currently being inferred.

[0111] The network node 1 820 maintains a list of ML models supported by the UE connected to it and a list of ML models supported by neighboring network nodes that can provide services to the UE in dual connectivity mode (DC).

[0112] Network node 1 820 checks whether ML model 1 currently performing inference with UE 810 is compatible with network node 2 822. Network node 1 820 may determine or check 836 that ML model 1 is compatible with network node 2 822. Network node 1 820 indicates 838 ML model 1 to network node 2 822.

[0113] UE 810 starts 840 inference with ML Model 1 of Network Node 1 and Network Node 2 .

[0114] Therefore, network node 1 820 ensures that the UE does not need to run different ML models for the same function for connections with different network nodes, thereby saving UE resources.

[0115] Fig. 9 An example of signaling between entities is shown. UE 910 is performing reasoning 930 with ML model 1 with network node 1 920, which may be a first network node or a master node (MN).

[0116] The UE 910 establishes 932 an auxiliary connection with a network node 2 922, which may be a second network node or an auxiliary node (SN). The UE 910 transmits information about at least one two-sided ML model supported by the UE 910 to the network node 2 922. The network node 2 922 may maintain a list of ML models supported by the UEs connected to it.

[0117] Network node 2 922 requests 934 from network node 1 information about the ML model for UE 910. For example, network node 2 922 may request an ML model ID for UE 910, such as the ML model ID currently used for UE 910.

[0118] If the UE 910 has already started inference with the network node 1 920 before establishing the auxiliary connection with the network node 2 922, then when indicating its ML capabilities at step 932, the UE 910 may indicate to the network node 2 922 the ML model currently being inferred.

[0119] Network node 1 920 maintains a list of ML models supported by UEs connected to it and a list of ML models supported by neighboring network nodes that can serve the UEs in DC mode.

[0120] Network node 1 920 checks whether ML model 1 currently inferring with UE 910 is compatible with network node 2 922. Network node 1 920 may determine 936 that the first two-sided machine learning model is not compatible with network node 2 922. Network node 1 920 selects 938 ML model 2 that is compatible with user equipment and network node 2 922.

[0121] Network Node 1 920 transmits 940 an instruction to switch to ML Model 2 to UE 910. Network Node 1 920 transmits an indication of ML Model 2 to Network Node 2.

[0122] UE 910 starts reasoning with the common model of Network Node 1 and Network Node 2 (ie, ML Model 2).

[0123] Therefore, network node 1 920 implements model coordination of UEs in DC mode.

[0124] If the network node does not support ML-related coordination, the UE can ensure model coordination in DC mode by itself. Fig.10An example of signaling between entities is shown. UE 1010 performs inference 1030 with network node 1 1020 using ML model 1. UE 1010 establishes 1032 an auxiliary connection with network node 2 1022. UE indicates its ML capabilities to network node 2 1022. Network node 2 1022 may maintain a list of ML models supported by UEs connected to it.

[0125] The UE 1010 receives 1034 an indication of the selected ML Model 2 from the network node 2 1022. The indication may be received when the UE 1010 performs inference of the ML Model 1 with the network node 2 1020.

[0126] UE 1010 transmits a NACK and an indication that the UE is inferring with another ML model (eg, ML Model 1) with Network Node 1 to Network Node 2. Network Node 1 is a Master Node (MN) and Network Node 2 is a Secondary Node (SN).

[0127] If UE 1010 has already started inference with network node 1 1020 before establishing the auxiliary connection with network node 2 1022, then when indicating its ML capabilities at step 1032, UE 1010 may indicate to network node 2 1022 the ML model currently being inferred.

[0128] The network node 2 1022 receives an indication of the ML model currently running at the UE. The network node 2 may accept 1038 the inference of the ML model 1 and transmit 1040 a positive acknowledgement (ACK). The network node 2 may reject 1038 the inference of the ML model 1 and transmit 1040 a negative acknowledgement (NACK). The acceptance or rejection may depend, for example, on the ML capabilities of the network node 2, or on the resources of the network node 2, such as available central processing unit (CPU) resources and / or occupancy of a hardware accelerator for ML.

[0129] Therefore, a UE in DC mode can indicate the ML model for the primary cell group (MCG) to the secondary cell group (SCG). This enables the secondary node to decide whether to run the same ML model that the UE has inferred.

[0130] Back to the O&M server, it understands the network topology and can ensure that neighboring network nodes have at least one compatible ML model for each ML-enabled function. Compatible ML models mean that if the UE is in dual connectivity mode and uses the same part of the bilateral ML model on both network nodes, then the ML models on these network nodes are compatible. For example, for CSI compression using an autoencoder, a UE in DC mode can run the same encoder model for two network nodes, while these network nodes can have different models for the decoder. Different models provide the same or sufficiently similar output for the same input. The process of configuring compatible ML models on neighboring network nodes can be driven by the O&M server.

[0131] Fig.11 An example of signaling between entities is shown. A network node 1110 joins 1130 a network managed by a management server (eg, O&M server 1120). When a new network node joins the network, a process of acquiring and configuring a compatible ML model may be triggered.

[0132] The O&M server 1120 identifies 1132 neighboring network nodes that may participate in the DC for the UE served by the network node 1110 .

[0133] The O&M server 1120 determines 1134 ML models that are compatible with the neighboring network node. The server 1120 may maintain a list of compatible ML models. The list may include a [neighbor ID; supported ML model ID] pair. The neighbor ID may include a New Radio Cell Global Identifier (NCGI) of neighbor 1, NCGI of neighbor 2, etc. The IDs of the supported ML models may be, for example, ID1, ID2, ID3 of neighbor 1; ID3, ID5, ID6 of neighbor 2.

[0134] Optionally, the O&M server 1120 may train 1134 one or more models that are compatible with neighboring network nodes.

[0135] The O&M server 1120 may configure an ML model compatible with neighboring network nodes for the network node 1110. For example, the O&M server 1120 may provide the ML model to the network node 1110. Therefore, the O&M server may be configured to provide a common ML model or information about a common ML model for a primary cell group and a secondary cell group serving a UE in a dual connectivity mode.

[0136] Fig.12 An example of a block diagram of an apparatus capable of performing at least the methods disclosed herein is shown. An apparatus 1200 is shown, which may include, for example, a mobile communication device, such as Figure 5 UE 510 in, or Figure 5The network node 520 in the device 1200. The device 1200 includes a processor 1210, which may include, for example, a single-core or multi-core processor, wherein the single-core processor includes one processing core and the multi-core processor includes multiple processing cores. In general, the processor 1210 may include a control device. The processor 1210 may include multiple processors. The processor 1210 may be a control device. The processing core may include, for example, a Cortex-A8 processing core manufactured by ARM Holdings, or a Steamroller processing core designed by Advanced Micro Devices Corporation. The processor 1210 may include at least one Qualcomm Snapdragon and / or Intel Atom processor. The processor 1210 may include at least one application-specific integrated circuit (ASIC). The processor 1210 may include at least one field programmable gate array (FPGA). The processor 1210 may be a device for performing method steps in the device 1200. The processor 1210 may be configured to perform operations at least in part by computer instructions.

[0137] The processor may include circuits, or be configured as circuits or groups of circuits, which are configured to perform the various stages of the method according to the example embodiments described herein. As used in this application, the term "circuit" may refer to one or more or all of the following: (a) a hardware-only circuit implementation (such as an implementation in analog and / or digital circuits only); and (b) a combination of hardware circuits and software, such as (as applicable): (i) a combination of analog and / or digital hardware circuits and software / firmware, and (ii) a hardware processor and any portion of software (including a digital signal processor), software and memory, which work together to enable a device such as a mobile phone or server to perform various functions; (c) a hardware circuit and / or processor that requires software (e.g., firmware) to operate, such as a microprocessor or a portion of a microprocessor, but the software may not be present when it is not required to operate.

[0138] This definition of circuitry applies to one or all uses of the term in this application, including in any claims. As a further example, as used in this application, the term "circuitry" also covers an implementation of only a hardware circuit or processor (or multiple processors) or a portion of a hardware circuit or processor and its (or their) accompanying software and / or firmware. For example, if applicable to a particular claim element, the term "circuitry" also covers a baseband integrated circuit or processor integrated circuit for a mobile device, or a similar integrated circuit in a server, cellular network device, or other computing or network device.

[0139] Device 1200 may include memory 1220. Memory 1220 may include random access memory and / or permanent memory. Memory 1220 may include at least one RAM chip. Memory 1220 may include, for example, solid-state, magnetic, optical and / or holographic memory. Memory 1220 may be at least partially accessible to processor 1210. Memory 1220 may be at least partially contained in processor 1210. Memory 1220 may be a device for storing information. Memory 1220 may include instructions, such as computer instructions or computer program code, and processor 1210 is configured to execute these instructions. When instructions configured to cause processor 1210 to perform certain operations are stored in memory 1220, and device 1200 as a whole is configured to run under the direction of processor 1210 using instructions from memory 1220, processor 1210 and / or at least one processing core thereof may be considered to be configured to perform the certain operations. Memory 1220 may be at least partially located outside device 1200 but accessible to device 1200.

[0140] Device 1200 may include a transmitter 1230. Device 1200 may include a receiver 1240. Transmitter 1230 and receiver 1240 may be configured to transmit and receive information according to at least one cellular or non-cellular standard, respectively. Transmitter 1230 may include multiple transmitters. Receiver 1240 may include multiple receivers. Transmitter 1230 and / or receiver 1240 may be configured to operate according to, for example, Global System for Mobile Communications (GSM), Wideband Code Division Multiple Access (WCDMA), 5G, Long Term Evolution (LTE), IS--95, Wireless Local Area Network (WLAN), Ethernet, and / or Worldwide Interoperability for Microwave Access (WiMAX) standards.

[0141] Device 1200 may include a user interface (UI) 1260. UI 1260 may include at least one of: a display, a keyboard, a touch screen, a vibrator configured to send a signal to the user by vibrating device 1200, a speaker, and a microphone. The user may operate device 1200 through UI 1260, such as answering an incoming call, making a phone call or video call, browsing the Internet, managing digital files stored in memory 1220, or digital files in the cloud accessible through transmitter 1230 and receiver 1240, or through NFC transceiver 1250, and / or playing games.

[0142] The processor 1210 may be equipped with a transmitter configured to output information from the processor 1210 to other devices contained in the device 1200 via electrical leads inside the device 1200. Such a transmitter may include a serial bus transmitter configured to output information to the memory 1220 for storage therein, for example, via at least one electrical lead. As an alternative to the serial bus, the transmitter may include a parallel bus transmitter. Similarly, the processor 1210 may include a receiver configured to receive information from other devices contained in the device 1200 via electrical leads inside the device 1200. Such a receiver may include a serial bus receiver configured to receive information from the receiver 1240, for example, via at least one electrical lead for processing in the processor 1210. As an alternative to the serial bus, the receiver may include a parallel bus receiver.

[0143] The term "non-transitory" as used herein is a restriction to the medium itself (ie, tangible, not a signal), not to the persistence of data storage (eg, RAM vs. ROM).

[0144] As used herein, “at least one of: ” and “at least one of ” and similar expressions (wherein a list of two or more elements is connected by “and” or “or”) refer to at least any one element, or at least any two or more elements, or at least all elements.

Claims

1. An apparatus comprising means for performing: transmitting information about at least one two-sided machine learning model supported by the apparatus to a network node; or transmitting a machine learning profile identification of the apparatus to a network node, wherein the machine learning profile identification is associated with at least one two-sided machine learning model supported by the apparatus; in, The at least one two-sided machine learning model is configured to enable the device and the network node to perform joint reasoning; An indication of at least one first selected two-sided machine learning model supported by the device is received from the network node.

2. The device according to claim 1, wherein: The apparatus is also configured to perform: Starting joint inference with the network nodes using the first selected two-sided machine learning model.

3. The device according to claim 1 or 2, wherein: The information about at least one two-sided machine learning model supported by the device includes: an identification of at least one two-sided machine learning model supported by the apparatus; and Information about the functionality of at least one two-sided machine learning model supported by the device.

4. A device as claimed in any one of the preceding claims, wherein: The apparatus is also configured to perform: Receiving an instruction from the network node to switch to a second two-sided machine learning model supported by the device, the second two-sided machine learning model also supported by another device served by the network node.

5. The device according to claim 4, wherein: The apparatus is also configured to perform: Switching to the second two-sided machine learning model; and Joint inference with the network nodes using the second two-sided machine learning model is started.

6. The device according to claim 4, wherein: The apparatus is also configured to perform: determining that the device no longer supports the second two-sided machine learning model; based on the determination, transmitting a negative acknowledgement to the network node; Continue with the joint inference of the two-sided machine learning model selected above.

7. A device as claimed in any one of the preceding claims, wherein: The apparatus is also configured to perform: Establishing a secondary connection with another network node; transmitting, to the other network node, information about at least one two-sided machine learning model supported by the apparatus; or transmitting a machine learning profile identification of the apparatus to the other network node, wherein the machine learning profile identification is associated with at least one two-sided machine learning model supported by the apparatus; Starting joint inference with the other network node using the first selected two-sided machine learning model.

8. The device according to claim 7, wherein: The apparatus is also configured to perform: receiving, from the network node, an instruction to switch to a second two-sided machine learning model supported by the device, the second two-sided machine learning model also supported by the other network node; Switch to the second two-sided machine learning model; Starting joint inference with the network node and the another network node using the second two-sided machine learning model.

9. The device according to claim 7, wherein: The device is further configured to perform: When performing joint reasoning with the network node using the first selected two-sided machine learning model, receiving an indication of a second selected two-sided machine learning model from the other network; Transmitting a negative acknowledgement to the other network along with an indication that the device and the network node are to perform joint inference using the first selected two-sided machine learning model.

10. The device of claim 9, wherein: The apparatus is also configured to perform: Receiving from the other network node an indication to start joint inference using the first selected two-sided machine learning model or an indication to refuse joint inference using the first selected two-sided machine learning model.

11. An apparatus comprising means for performing: a) receiving, from a user device, information about at least one two-sided machine learning model supported by the user device; or b) receiving, from a user device, a machine learning profile identifier of the user device, wherein the machine learning profile identifier is associated with at least one two-sided machine learning model supported by the user device; and acquiring information about at least one bilateral machine learning model supported by the user device based on a machine learning profile identifier of the user device; in, The at least one two-sided machine learning model is configured to enable joint reasoning by the device and the user device; Selecting a first two-sided machine learning model supported by the user device; Transmitting to the user device an indication of at least selecting the first two-sided machine learning model supported by the user device.

12. The device of claim 11, wherein: The apparatus is also configured to perform: Start joint inference with the user device using the first two-sided machine learning model.

13. The device according to claim 11 or 12, wherein: The first two-sided machine learning model is selected based on other two-sided machine learning models that the device and one or more other user devices are currently jointly inferring or that are supported by the one or more other user devices.

14. The device according to any one of claims 11 to 13, wherein: The apparatus is also configured to perform: a) receiving, from another user device served by the apparatus, information about at least one two-sided machine learning model supported by the other user device; or b) receiving, from another user device served by the apparatus, a machine learning profile identifier of the other user device, wherein the machine learning profile identifier is associated with at least one two-sided machine learning model supported by the other user device; and Information about at least one two-sided machine learning model supported by the other user device is obtained based on the machine learning profile identifier of the other user device.

15. The device of claim 14, wherein: The apparatus is also configured to perform: Determining a second bilateral machine learning model supported by both user devices based on information about at least one bilateral machine learning model supported by the user device and the other user device; Transmit an instruction to the user device to switch to the second two-sided machine learning model.

16. The device of claim 15, wherein: The apparatus is also configured to perform: transmitting an indication of selecting the second two-sided machine learning model to the other user equipment; and Start joint inference with the user device and the other user device using a second two-sided machine learning model.

17. The device according to claim 15 or 16, wherein: The apparatus is also configured to perform: receiving a negative acknowledgement from the user equipment indicating that the user equipment no longer supports the second two-sided machine learning model; Continue joint reasoning with the user device using the first two-sided machine learning model; and Start joint inference with the other user device using a second two-sided machine learning model.

18. The device according to any one of claims 11 to 17, wherein: The apparatus is also configured to perform: receiving, from another network node serving the user equipment, a request for information about a two-sided machine learning model for the user equipment; determining that the first two-sided machine learning model is compatible with capabilities of the another network node; An indication of selecting the first two-sided machine learning model is transmitted to the other network node.

19. The device according to any one of claims 11 to 17, wherein: The apparatus is also configured to perform: receiving, from another network node serving the user equipment, a request for information about a two-sided machine learning model for the user equipment; determining that the first two-sided machine learning model is incompatible with capabilities of the another network node; Based on the determination, selecting a second two-sided machine learning model that is compatible with capabilities of the user equipment and capabilities of the another network node; Transmit an instruction to the user device to switch to the second two-sided machine learning model.

20. The device of claim 19, wherein: The apparatus is also configured to perform: transmitting an indication of selecting the second two-sided machine learning model to the other network node; and Initiate inference with the user equipment and the another network node using the second two-sided machine learning model.

21. An apparatus comprising means for performing the following operations: Establishing a connection with a user equipment, the apparatus being configured to serve the user equipment as an auxiliary network node; receiving, from a primary network node configured to serve the user equipment, information about a two-sided machine learning model that is compatible with capabilities of the user equipment and the secondary network node; Start joint inference with the user device and the master network node using a two-sided machine learning model.

22. A device as claimed in any preceding claim, wherein: The device includes at least one processor; and at least one memory storing instructions, and when the instructions are executed by the at least one processor, the device performs corresponding operations.

23. A method comprising: transmitting, by a device to a network node, information about at least one two-sided machine learning model supported by the device; or transmitting, by a device, a machine learning profile identifier of the device to a network node, wherein the machine learning profile identifier is associated with at least one two-sided machine learning model supported by the device; wherein the at least one two-sided machine learning model is configured to enable the device and the network node to perform joint reasoning; An indication of at least one first selected two-sided machine learning model supported by the device is received from the network node.

24. A method comprising: a) receiving, by a network node, from a user equipment, information about at least one two-sided machine learning model supported by the user equipment; or b) receiving, by a network node, a machine learning profile identifier of a user equipment from the user equipment, wherein the machine learning profile identifier is associated with at least one two-sided machine learning model supported by the user equipment; and acquiring information about at least one bilateral machine learning model supported by the user device based on a machine learning profile identifier of the user device; wherein the at least one two-sided machine learning model is configured to enable the apparatus and the user equipment to perform joint reasoning; The network node selects a first two-sided machine learning model supported by the user equipment; An indication is transmitted by the network node to the user equipment to select at least the first two-sided machine learning model supported by the user equipment.

25. A method comprising: Establishing a connection with a user equipment by a device, the device being configured to serve the user equipment as an auxiliary network node; receiving, by the apparatus, from a primary network node configured to serve the user equipment, information about a two-sided machine learning model that is compatible with capabilities of the user equipment and the secondary network node; The device initiates joint inference with the user equipment and the master network node using a two-sided machine learning model.

26. A computer program comprising instructions which, when executed by an apparatus, cause the apparatus to: transmitting information about at least one two-sided machine learning model supported by the apparatus to a network node; or transmitting a machine learning profile identifier of the device to a network node, wherein: The machine learning profile identifier is associated with at least one two-sided machine learning model supported by the device; wherein the at least one two-sided machine learning model is configured to enable the device and the network node to perform joint reasoning; An indication of at least one first selected two-sided machine learning model supported by the device is received from the network node.

27. A computer program comprising instructions which, when executed by an apparatus, cause the apparatus to: a) receiving, from a user device, information about at least one two-sided machine learning model supported by the user device; or b) receiving a machine learning profile identifier of the user equipment from the user equipment, wherein: The machine learning profile identifier is associated with at least one two-sided machine learning model supported by the user device; and acquiring information about at least one bilateral machine learning model supported by the user device based on a machine learning profile identifier of the user device; wherein the at least one two-sided machine learning model is configured to enable the apparatus and the user equipment to perform joint reasoning; Selecting a first two-sided machine learning model supported by the user device; Transmitting to the user device an indication of at least selecting the first two-sided machine learning model supported by the user device.

28. A computer program comprising instructions which, when executed by an apparatus, cause the apparatus to: Establishing a connection with a user equipment, the apparatus being configured to serve the user equipment as an auxiliary network node; receiving, from a primary network node configured to serve the user equipment, information about a two-sided machine learning model that is compatible with capabilities of the user equipment and the secondary network node; Start joint inference with the user device and the master network node using a two-sided machine learning model.

29. A non-transitory computer-readable medium comprising instructions that, when executed by an apparatus, cause the apparatus to at least perform: transmitting information about at least one two-sided machine learning model supported by the apparatus to a network node; or transmitting a machine learning profile identifier of the device to a network node, wherein: The machine learning profile identifier is associated with at least one two-sided machine learning model supported by the device; wherein the at least one two-sided machine learning model is configured to enable the device and the network node to perform joint reasoning; An indication of at least one first selected two-sided machine learning model supported by the device is received from the network node.

30. A non-transitory computer-readable medium comprising instructions that, when executed by an apparatus, cause the apparatus to at least perform: Receiving, from a user device, information about at least one two-sided machine learning model supported by the user device; or Receiving, from a user device, a machine learning profile identifier of the user device, wherein: The machine learning profile identifier is associated with at least one two-sided machine learning model supported by the user device; and acquiring information about at least one bilateral machine learning model supported by the user device based on a machine learning profile identifier of the user device; wherein the at least one two-sided machine learning model is configured to enable the apparatus and the user equipment to perform joint reasoning; Selecting a first two-sided machine learning model supported by the user device; Transmitting to the user device an indication of at least selecting the first two-sided machine learning model supported by the user device.

31. A non-transitory computer-readable medium comprising instructions that, when executed by an apparatus, cause the apparatus to at least perform: Establishing a connection with a user equipment, the apparatus being configured to serve the user equipment as an auxiliary network node; receiving, from a primary network node configured to serve the user equipment, information about a two-sided machine learning model that is compatible with capabilities of the user equipment and the secondary network node; Start joint inference with the user device and the master network node using a two-sided machine learning model.

Images