Federal parameter training for machine learning

Abstract

A UE receives parameters of a machine learning model from a network node; a gradient value is calculated relative to one of the parameters of the machine learning model, the gradient value including a positive gradient value or a negative gradient value. The UE may send, in one of a pair of Resource Elements (REs), an analog signal to the network node indicating a magnitude of the gradient value relative to the one of the parameters of the machine learning model. The UE may transmit an analog signal to the network node or skip transmission of the analog signal in a single RE designated to indicate the positive gradient value or the negative gradient value relative to the gradient value of the one of the parameters of the machine learning model based on the gradient value.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the benefit of Greek patent application serial number 20220100426, filed on May 23, 2022, entitled “FEDERATED PARAMETER TRAINING FOR MACHINE LEARNING,” which is expressly incorporated herein by reference in its entirety. Technical Field

[0003] The present disclosure relates generally to communication systems and, more particularly, to a method of wireless communication including federated parameter training. Background Art

[0004] Wireless communication systems are widely deployed to provide a variety of telecommunication services, such as telephony, video, data, messaging, and broadcast. Typical wireless communication systems may employ multiple access technologies capable of supporting communication with multiple users by sharing available system resources. Examples of such multiple access technologies include code division multiple access (CDMA) systems, time division multiple access (TDMA) systems, frequency division multiple access (FDMA) systems, orthogonal frequency division multiple access (OFDMA) systems, single carrier frequency division multiple access (SC-FDMA) systems, and time division synchronous code division multiple access (TD-SCDMA) systems.

[0005] These multiple access technologies have been adopted in various telecommunication standards to provide a common protocol that enables different wireless devices to communicate at a city, country, region, or even global level. An example telecommunication standard is 5G New Radio (NR). 5G NR is part of the continuous mobile broadband evolution promulgated by the Third Generation Partnership Project (3GPP) to meet new requirements associated with latency, reliability, security, scalability (e.g., with the Internet of Things (IoT)) and other requirements. 5GNR includes services associated with enhanced mobile broadband (eMBB), massive machine type communications (mMTC), and ultra-reliable low latency communications (URLLC). Certain aspects of 5G NR may be based on the 4G Long Term Evolution (LTE) standard. Further improvements to 5G NR technology are needed. In addition, these improvements may also be applicable to other multiple access technologies and telecommunication standards that employ these technologies. Summary of the invention

[0006] A simplified summary of one or more aspects is presented below to provide a basic understanding of these aspects. This summary is not an extensive overview of all contemplated aspects. This summary is neither intended to identify key or important elements of all aspects, nor to describe the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as a prelude to a more detailed description presented later.

[0007] In one aspect of the present disclosure, a method, computer-readable medium, and apparatus for wireless communication at a user equipment (UE) are provided. The apparatus receives parameters of a machine learning model from a network node. The apparatus calculates a gradient value relative to one of the parameters of the machine learning model, the gradient value comprising a positive gradient value or a negative gradient value. The apparatus sends an analog signal indicating the magnitude of the gradient value relative to the parameter in the parameters of the machine learning model to the network node in one of a pair of resource elements (REs), the pair of REs being designated for indicating the gradient value relative to the parameter in the parameters of the machine learning model.

[0008] In one aspect of the present disclosure, a method, computer-readable medium, and apparatus for wireless communication at a user equipment (UE) are provided. The apparatus receives parameters of a machine learning model from a network node. The apparatus calculates a gradient value relative to one of the parameters of the machine learning model, the sign of the gradient value being one of positive or negative. The apparatus sends a simulated signal to the network node in a single RE or skips sending the simulated signal based on the sign of the gradient value relative to the parameter in the parameters of the machine learning model, the single RE being designated for indicating the sign of the gradient value relative to the parameter in the parameters of the machine learning model.

[0009] In one aspect of the present disclosure, a method, a computer-readable medium, and a device for wireless communication at a network node are provided. The device outputs parameters of a machine learning model for multiple UEs for transmission. The device obtains at least one aggregated analog signal in a pair of REs designated for the multiple UEs to indicate a gradient value of one parameter in the parameters of the machine learning model relative to the multiple UEs, the at least one aggregated analog signal obtained in the pair of REs representing multiple analog signals cumulatively obtained from the multiple UEs at the network node, each of the multiple analog signals indicating the gradient value of the parameter in the parameters of the machine learning model calculated at the corresponding UE in the multiple UEs, the gradient value including a positive gradient value or a negative gradient value. The device updates the parameter in the parameters of the machine learning model based on the at least one aggregated analog signal obtained in the pair of REs.

[0010] In one aspect of the present disclosure, a method, a computer-readable medium, and an apparatus for wireless communication at a network node are provided. The apparatus outputs parameters of a machine learning model for multiple UEs for transmission. The apparatus obtains an aggregated analog signal in a single RE designated for the multiple UEs to indicate a sign of a gradient value of one of the parameters of the machine learning model relative to the multiple UEs, the aggregated analog signal obtained in the single RE representing multiple analog signals cumulatively obtained from the multiple UEs at the network node, each of the multiple analog signals indicating that the sign of the gradient value of the parameter in the parameters of the machine learning model calculated at the corresponding UE in the multiple UEs is one of positive or negative. The apparatus updates the parameter in the parameters of the machine learning model based on the aggregated analog signal obtained in the single RE.

[0011] To achieve the aforementioned and related purposes, one or more aspects include the features fully described below and particularly pointed out in the claims. The following description and the accompanying drawings set forth in detail some exemplary features of one or more aspects. However, these features are only indicative of only some of the various ways in which the principles of the various aspects can be employed. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 is a diagram illustrating an example of a wireless communication system and an access network.

[0013] Figure 2A is a diagram illustrating an example of a first frame according to various aspects of the present disclosure.

[0014] Figure 2B is a diagram illustrating an example of DL channels within a subframe according to various aspects of the present disclosure.

[0015] Figure 2C is a diagram illustrating an example of a second frame according to various aspects of the present disclosure.

[0016] Figure 2D is a diagram illustrating an example of UL channels within a subframe according to various aspects of the present disclosure.

[0017] Figure 3 is a diagram illustrating an example of a base station and a user equipment (UE) in an access network.

[0018] Figure 4 is an example of an AI / ML algorithm for a method of wireless communication.

[0019] Figure 5 is an example of an ML parameter indicating a method of wireless communication.

[0020] Figure 6is an example of an ML parameter indicating a method of wireless communication.

[0021] Figure 7 is an example of an ML parameter indicating a method of wireless communication.

[0022] Figure 8 is a call flow diagram of a method of wireless communication.

[0023] Fig. 9 is a flow chart of a method of wireless communication.

[0024] Fig.10 is a flow chart of a method of wireless communication.

[0025] Fig.11 is a flow chart of a method of wireless communication.

[0026] Fig.12 is a flow chart of a method of wireless communication.

[0027] Fig.13 is a flow chart of a method of wireless communication.

[0028] Fig.14 is a flow chart of a method of wireless communication.

[0029] Fig.15 is a flow chart of a method of wireless communication.

[0030] Fig.16 is a flow chart of a method of wireless communication.

[0031] Fig.17 are diagrams illustrating examples of hardware implementations for example apparatuses and / or network entities.

[0032] Fig.18 is a diagram illustrating an example of a hardware implementation for an example network entity.

[0033] Fig.19 is a diagram illustrating an example of a hardware implementation for an example network entity. DETAILED DESCRIPTION

[0034] The machine learning model may include tens, hundreds, thousands or more parameters. In some aspects, multiple UEs may contribute to a federated learning training session. However, sending local gradients to a server may generate a lot of overhead. In some aspects, air aggregation can be used to obtain the sum of gradient values ​​from multiple UEs, which can reduce the overhead of federated learning. The various aspects presented herein provide an air aggregation mechanism that is robust to phase uncertainty (e.g., insensitive to phase uncertainty). Reduced sensitivity to phase uncertainty allows relaxation of phase synchronization between devices, which can reduce the complexity of federated learning. The various aspects presented herein provide a non-coherent air aggregation mechanism that allows gradient reporting in a more efficient and / or more efficient manner. In some aspects, a UE may use a pair of REs to report a gradient value (e.g., the absolute magnitude of the gradient). The reporting of this magnitude may allow the energy received at the network to be proportional to the number of UEs that calculate negative or positive values, and may provide additional confidence for the gradient sign determined by the network. In some aspects, a UE may use switch signaling to indicate a positive or negative gradient using a single RE, which may further reduce the overhead of gradient reporting.

[0035] The detailed description set forth below in conjunction with the accompanying drawings is a description of various configurations and does not represent the only configurations with which the concepts described herein can be practiced. In order to provide a thorough understanding of the various concepts, the specific embodiments include specific details. However, these concepts can be practiced without these specific details. In some instances, well-known structures and components are shown in block diagram form to avoid blurring these concepts.

[0036] Several aspects of the telecommunication system are also presented with reference to various apparatuses and methods. These apparatuses and methods are described in the following detailed description and illustrated in the accompanying drawings by various blocks, components, circuits, processes, algorithms, etc. (collectively referred to as "elements"). These elements may be implemented using electronic hardware, computer software, or any combination thereof. Whether these elements are implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system.

[0037] As an example, an element, or any part of an element, or any combination of elements, may be implemented as a "processing system" comprising one or more processors. Examples of processors include microprocessors, microcontrollers, graphics processing units (GPUs), central processing units (CPUs), application processors, digital signal processors (DSPs), reduced instruction set computing (RISC) processors, systems on chip (SoCs), baseband processors, field programmable gate arrays (FPGAs), programmable logic devices (PLDs), state machines, gate logic, discrete hardware circuits, and other suitable hardware configured to perform various functionalities described throughout the present disclosure. One or more processors in a processing system may execute software. Whether referred to as software, firmware, middleware, microcode, hardware description language, or other terms, software should be broadly interpreted as meaning instructions, instruction sets, codes, code segments, program codes, programs, subroutines, software components, applications, software applications, software packages, routines, subroutines, objects, executables, threads of execution, procedures, functions, or any combination thereof.

[0038] Thus, in one or more example aspects, specific implementations and / or use cases, the described functionality may be implemented in hardware, software, or any combination thereof. If implemented in software, the functionality may be stored or encoded as one or more instructions or codes on a computer-readable medium. Computer-readable media include computer storage media. Storage media may be any available media that can be accessed by a computer. As an example, such computer-readable media may include random access memory (RAM), read-only memory (ROM), electrically erasable programmable ROM (EEPROM), optical disk storage, magnetic disk storage, other magnetic storage devices, combinations of these types of computer-readable media, or any other medium that can be used to store computer-executable code in the form of instructions or data structures that can be accessed by a computer.

[0039] Although various aspects, specific implementations and / or use cases are described in this application by the illustration of some examples, additional or different aspects, specific implementations and use cases may be generated in many different arrangements and scenarios. The various aspects, specific implementations and / or use cases described herein can be implemented across many different platform types, devices, systems, shapes, sizes and packaging arrangements. For example, various aspects, specific implementations and / or use cases can be generated via integrated chip specific implementations and other devices based on non-module components (e.g., end-user devices, vehicles, communication equipment, computing equipment, industrial equipment, retail / purchase equipment, medical equipment, devices that enable artificial intelligence (AI), etc.). Although some examples may or may not be specifically for use cases or applications, the described examples may have a wide range of applicability. Various aspects, specific implementations and / or use cases can be in the range from chip-level or modular components to non-modular, non-chip-level specific implementations, and further to the range of aggregated, distributed or original equipment manufacturer (OEM) devices or systems in conjunction with one or more technologies herein. In some practical environments, the equipment in conjunction with the various aspects and features described may also include additional components and features for implementing and practicing the claimed and described aspects. For example, the transmission and reception of wireless signals necessarily include multiple components for analog and digital purposes (e.g., hardware components including antennas, RF chains, power amplifiers, modulators, buffers, processors, interleavers, adders / summers, etc.). The techniques described herein can be practiced in a wide variety of devices of various sizes, shapes, and configurations, chip-level components, systems, distributed arrangements, aggregated or disaggregated components, end-user devices, etc.

[0040] The deployment of a communication system (such as a 5G NR system) can be arranged with various components or components in a variety of ways. In a 5G NR system or network, a network node, a network entity, a mobility element of a network, a radio access network (RAN) node, a core network node, a network element or network equipment (such as a base station (BS)), or one or more units (or one or more components) performing base station functionality can be implemented in an aggregated or decomposed architecture. For example, a BS (such as a Node B (NB), an evolved NB (eNB), an NR BS, a 5G NB, an access point (AP), a transmit receive point (TRP) or a cell, etc.) can be implemented as an aggregated base station (also referred to as an independent BS or a monolithic BS) or a decomposed base station.

[0041] A converged base station may be configured to utilize a radio protocol stack physically or logically integrated within a single RAN node. A decomposed base station may be configured to utilize a protocol stack physically or logically distributed between two or more units, such as one or more central or centralized units (CUs), one or more distributed units (DUs), or one or more radio units (RUs). In some aspects, a CU may be implemented within a RAN node, and one or more DUs may be co-located with the CU, or alternatively, may be geographically or virtually distributed in one or more other RAN nodes. A DU may be implemented to communicate with one or more RUs. Each of a CU, a DU, and a RU may be implemented as a virtual unit, i.e., a virtual central unit (VCU), a virtual distributed unit (VDU), or a virtual radio unit (VRU).

[0042] Base station operations or network design may take into account the aggregated nature of base station functionality. For example, a disaggregated base station may be utilized in an integrated access backhaul (IAB) network, an open radio access network (O-RAN (network configuration such as that initiated by the O-RAN Alliance)), or a virtualized radio access network (vRAN, also known as a cloud radio access network (C-RAN)). Decomposition may include distributing functionality across two or more units at various physical locations, as well as virtually distributing functionality of at least one unit, which may enable flexibility in network design. Individual units of a disaggregated base station or disaggregated RAN architecture may be configured for wired or wireless communication with at least one other unit.

[0043] Figure 1 1 is a diagram 100 illustrating an example of a wireless communication system and an access network. The illustrated wireless communication system includes a disaggregated base station architecture. The disaggregated base station architecture may include one or more CUs 110, which may communicate directly with a core network 120 via a backhaul link, or indirectly with the core network 120 through one or more disaggregated base station units (such as a near real-time (near RT) RAN intelligent controller (RIC) 125 via an E2 link, or a non-real-time (non-RT) RIC 115 associated with a service management and orchestration (SMO) framework 105, or both). The CU 110 may communicate with one or more DUs 130 via corresponding midhaul links (such as an F1 interface). The DU 130 may communicate with one or more RUs 140 via corresponding fronthaul links. The RU 140 may communicate with corresponding UEs 104 via one or more radio frequency (RF) access links. In some implementations, a UE 104 may be served simultaneously by multiple RUs 140.

[0044] Each of the units (i.e., CU 110, DU 130, RU 140, and near-RT RIC 125, non-RT RIC 115, and SMO framework 105) may include or be coupled to one or more interfaces configured to receive or send signals, data, or information (collectively, signals) via a wired or wireless transmission medium. Each of these units or an associated processor or controller that provides instructions to the communication interface of these units may be configured to communicate with one or more of the other units via a transmission medium. For example, these units may include a wired interface configured to receive signals or send signals to one or more of the other units via a wired transmission medium. Additionally, these units may include a wireless interface that may include a receiver, transmitter, or transceiver (such as an RF transceiver) configured to receive signals or send signals to one or more of the other units via a wireless transmission medium, or both.

[0045] In some aspects, CU 110 may host one or more higher layer control functions. Such control functions may include radio resource control (RRC), packet data convergence protocol (PDCP), or service data adaptation protocol (SDAP), etc. Each control function may be implemented using an interface that is configured to communicate signals with other control functions hosted by CU 110. CU 110 may be configured to handle user plane functionality (i.e., central unit-user plane (CU-UP)), control plane functionality (i.e., central unit-control plane (CU-CP)), or a combination thereof. In some specific implementations, CU 110 may be logically split into one or more CU-UP units and one or more CU-CP units. When implemented in an O-RAN configuration, the CU-UP unit may communicate bidirectionally with the CU-CP unit via an interface (such as an E1 interface). As needed, CU 110 may be implemented to communicate with DU 130 for network control and signaling.

[0046] DU 130 may correspond to a logical unit that includes one or more base station functions for controlling the operation of one or more RUs 140. In some aspects, DU 130 may host one or more of a radio link control (RLC) layer, a medium access control (MAC) layer, and one or more high physical (PHY) layers (such as modules for forward error correction (FEC) encoding and decoding, scrambling, modulation and demodulation, etc.) depending at least in part on a functional split such as that defined by 3GPP. In some aspects, DU 130 may also host one or more low PHY layers. Each layer (or module) may be implemented using an interface that is configured to communicate signals with other layers (and modules) hosted by DU 130 or with control functions hosted by CU 110.

[0047] The lower layer functionality may be implemented by one or more RUs 140. In some deployments, the RU 140 controlled by the DU 130 may correspond to a logical node that hosts RF processing functions or low PHY layer functions (such as performing fast Fourier transform (FFT), inverse FFT (iFFT), digital beamforming, or physical random access channel (PRACH) extraction and filtering, etc.) or both based at least in part on functional splitting (such as lower layer functional splitting). In such an architecture, the RU 140 may be implemented to handle over-the-air (OTA) communications with one or more UEs 104. In some specific implementations, real-time and non-real-time aspects of control plane communications and user plane communications with the RU 140 may be controlled by the corresponding DU 130. In some scenarios, this configuration may enable the DU 130 and the CU 110 to be implemented in a cloud-based RAN architecture (such as a vRAN architecture).

[0048] The SMO framework 105 may be configured to support RAN deployment and provisioning of non-virtualized network elements and virtualized network elements. For non-virtualized network elements, the SMO framework 105 may be configured to support the deployment of dedicated physical resources for RAN coverage requirements, which may be managed via an operation and maintenance interface (such as an O1 interface). For virtualized network elements, the SMO framework 105 may be configured to interact with a cloud computing platform (such as an open cloud (O-Cloud) 190) to perform network element lifecycle management (such as instantiating virtualized network elements) via a cloud computing platform interface (such as an O2 interface). Such virtualized network elements may include, but are not limited to, CU 110, DU 130, RU 140, and near-RT RIC 125. In some specific implementations, the SMO framework 105 may communicate with hardware aspects of the 4G RAN (such as an open eNB (O-eNB) 111) via the O1 interface. Additionally, in some specific implementations, the SMO framework 105 may communicate directly with one or more RUs 140 via the O1 interface. The SMO framework 105 may also include a non-RT RIC 115 configured to support the functionality of the SMO framework 105 .

[0049] The non-RT RIC 115 may be configured to include logic functions that enable non-real-time control and optimization of RAN elements and resources, artificial intelligence (AI) / machine learning (ML) (AI / ML) workflows including model training and updating, or policy-based guidance of applications / features in the near-RT RIC 125. The non-RT RIC 115 may be coupled to or communicate with the near-RT RIC 125 (such as via an A1 interface). The near-RT RIC 125 may be configured to include logic functions that enable near-real-time control and optimization of RAN elements and resources via data collection and actions through an interface (such as via an E2 interface) that connects one or more CUs 110, one or more DUs 130, or both, and the O-eNB with the near-RT RIC 125.

[0050] In some implementations, in order to generate an AI / ML model to be deployed in the near-RT RIC 125, the non-RT RIC 115 may receive parameters or external enrichment information from an external server. Such information may be utilized by the near-RT RIC 125 and may be received from a non-network data source or from a network function at the SMO framework 105 or the non-RT RIC 115. In some examples, the non-RT RIC 115 or the near-RT RIC 125 may be configured to tune RAN behavior or performance. For example, the non-RT RIC 115 may monitor long-term trends and patterns of performance and employ AI / ML models to perform corrective actions through the SMO framework 105 (such as via reconfiguration of O1) or via the creation of RAN management policies (such as A1 policies).

[0051] At least one of the CU 110, the DU 130, and the RU 140 may be referred to as a base station 102. Therefore, the base station 102 may include one or more of the CU 110, the DU 130, and the RU 140 (each component is indicated by a dotted line to indicate that each component may be included in the base station 102 or may not be included in the base station). The base station 102 provides an access point to the core network 120 for the UE 104. The base station 102 may include a macro cell (a high-power cellular base station) and / or a small cell (a low-power cellular base station). Small cells include femto cells, pico cells, and micro cells. A network including both small cells and macro cells may be referred to as a heterogeneous network. A heterogeneous network may also include a home evolved Node B (eNB) (HeNB), which may provide services to a restricted group referred to as a closed subscriber group (CSG). The communication link between RU 140 and UE 104 may include uplink (UL) (also known as reverse link) transmission from UE 104 to RU 140 and / or downlink (DL) (also known as forward link) transmission from RU 140 to UE 104. The communication link may use multiple-input multiple-output (MIMO) antenna technology, including spatial multiplexing, beamforming and / or transmission diversity. The communication link may pass through one or more carriers. For each carrier allocated in a carrier aggregation of up to Yx MHz (x component carriers) for transmission in each direction, the base station 102 / UE 104 may use a spectrum of up to Y MHz (e.g., 5MHz, 10MHz, 15MHz, 20MHz, 100MHz, 400MHz, etc.) bandwidth. The carriers may or may not be adjacent to each other. The allocation of carriers may be asymmetric with respect to DL and UL (e.g., more or fewer carriers may be allocated to DL compared to UL). The component carrier may include a primary component carrier and one or more secondary component carriers. The primary component carrier may be referred to as a primary cell (PCell) and the secondary component carrier may be referred to as a secondary cell (SCell).

[0052] Some UEs 104 may communicate with each other using a device-to-device (D2D) communication link 158. The D2D communication link 158 may use DL / UL wireless wide area network (WWAN) spectrum. The D2D communication link 158 may use one or more sidelink channels, such as a physical sidelink broadcast channel (PSBCH), a physical sidelink discovery channel (PSDCH), a physical sidelink shared channel (PSSCH), and a physical sidelink control channel (PSCCH). The D2D communication may be through various wireless D2D communication systems, such as, for example, Bluetooth, Wi-Fi based on the Institute of Electrical and Electronics Engineers (IEEE) 802.11 standard, LTE, or NR.

[0053] The wireless communication system may also include a Wi-Fi AP 150 that communicates with a UE 104 (also referred to as a Wi-Fi station (STA)) via a communication link 154, e.g., in a 5 GHz unlicensed spectrum, etc. When communicating in an unlicensed spectrum, the UE 104 / AP 150 may perform a clear channel assessment (CCA) prior to communication to determine whether a channel is available.

[0054] The electromagnetic spectrum is typically subdivided into various categories, bands, channels, etc. based on frequency / wavelength. In 5G NR, two initial operating bands have been identified as frequency range designations FR1 (410 MHz–7.125 GHz) and FR2 (24.25 GHz–52.6 GHz). Although a portion of FR1 is greater than 6 GHz, FR1 is often referred to (interchangeably) as the “sub-6 GHz” band in various documents and articles. A similar naming issue sometimes occurs with respect to FR2, which is often (interchangeably) referred to as the “millimeter wave” band in documents and articles, although it is different from the extremely high frequency (EHF) band (30 GHz–300 GHz) identified as the “millimeter wave” band by the International Telecommunication Union (ITU).

[0055] Frequencies between FR1 and FR2 are generally referred to as mid-band frequencies. Recent 5G NR research has identified the operating bands for these mid-band frequencies as frequency range designation FR3 (7.125 GHz–24.25 GHz). The bands falling within FR3 can inherit FR1 characteristics and / or FR2 characteristics, and therefore can effectively extend the features of FR1 and / or FR2 to mid-band frequencies. In addition, higher frequency bands are currently being explored to extend 5G NR operations to more than 52.6 GHz. For example, three higher operating bands have been identified as frequency range designations FR2-2 (52.6 GHz–71 GHz), FR4 (71 GHz–114.25 GHz), and FR5 (114.25 GHz–300 GHz). Each of these higher frequency bands falls within the EHF band.

[0056] In view of the above, unless otherwise specifically stated, if the term "below 6 GHz" or the like is used in this document, it may broadly refer to frequencies that may be less than 6 GHz, may be within FR1, or may include mid-band frequencies. In addition, unless otherwise specifically stated, if the term "millimeter wave" or the like is used in this document, it may broadly refer to frequencies that may include mid-band frequencies, may be within FR2, FR4, FR2-2, and / or FR5, or may be within the EHF band.

[0057] The base station 102 and the UE 104 may each include multiple antennas (such as antenna elements, antenna panels, and / or antenna arrays) to facilitate beamforming. The base station 102 may send a beamformed signal 182 to the UE 104 in one or more transmit directions. The UE 104 may receive the beamformed signal from the base station 102 in one or more receive directions. The UE 104 may also send a beamformed signal 184 to the base station 102 in one or more transmit directions. The base station 102 may receive the beamformed signal from the UE 104 in one or more receive directions. The base station 102 / UE 104 may perform beam training to determine the best receive direction and the best transmit direction for each of the base station 102 / UE 104. The transmit direction and the receive direction of the base station 102 may be the same or may not be the same. The transmit direction and the receive direction of the UE 104 may be the same or may not be the same.

[0058] The base station 102 may include and / or be referred to as a gNB, a Node B, an eNB, an access point, a base transceiver station, a radio base station, a radio transceiver, a transceiver function, a basic service set (BSS), an extended service set (ESS), a transmit receive point (TRP), a network node, a network entity, a network equipment, or some other suitable term. The base station 102 may be implemented as an integrated access and backhaul (IAB) node, a relay node, a sidelink node, an aggregated (monolithic) base station with a baseband unit (BBU) (including a CU and a DU) and a RU, or as a decomposed base station including one or more of a CU, a DU, and / or a RU. A set of base stations that may include decomposed base stations and / or aggregated base stations may be referred to as a next generation (NG) RAN (NG-RAN).

[0059] The core network 120 may include an access and mobility management function (AMF) 161, a session management function (SMF) 162, a user plane function (UPF) 163, a unified data management (UDM) 164, one or more location servers 168, and other functional entities. AMF 161 is a control node that handles signaling between UE 104 and the core network 120. AMF 161 supports registration management, connection management, mobility management, and other functions. SMF 162 supports session management and other functions. UPF 163 supports packet routing, packet forwarding, and other functions. UDM 164 supports the generation of authentication and key agreement (AKA) credentials, user identity handling, access authorization, and subscription management. One or more location servers 168 are illustrated as including a gateway mobile location center (GMLC) 165 and a location management function (LMF) 166. However, in general, the one or more location servers 168 may include one or more location / positioning servers, which may include one or more of the GMLC 165, LMF 166, positioning determination entity (PDE), serving mobile location center (SMLC) or mobile positioning center (MPC), etc. The GMLC 165 and LMF 166 support UE location services. The GMLC 165 provides an interface for clients / applications (e.g., emergency services) to access UE positioning information. The LMF 166 receives measurements and assistance information from the NG-RAN and UE 104 via the AMF 161 to calculate the positioning of the UE 104. The NG-RAN may utilize one or more positioning methods in order to determine the positioning of the UE 104. Positioning the UE 104 may involve signal measurements, position estimation, and optional speed calculations based on these measurements. Signal measurements may be performed by the UE 104 and / or the serving base station 102. The measured signals may be based on a satellite positioning system (SPS) 170 (e.g., one or more of a global navigation satellite system (GNSS), a global positioning system (GPS), a non-terrestrial network (NTN), or other satellite positioning / location systems), LTE signals, wireless local area network (WLAN) signals, Bluetooth signals, a terrestrial beacon system (TBS), sensor-based information (e.g., an atmospheric pressure sensor, a motion sensor), an NR enhanced cell ID (NR E-CID) method, NR signals (e.g., multi-round trip time (multi-RTT), DL angle of departure (DL-AoD), DL time difference of arrival (DL-TDOA), UL time difference of arrival (UL-TDOA), and UL angle of arrival (UL-AoA) positioning), and / or one or more of other systems / signals / sensors.

[0060] Examples of UE 104 include cellular phones, smart phones, session initiation protocol (SIP) phones, laptops, personal digital assistants (PDAs), satellite radios, global positioning systems, multimedia devices, video devices, digital audio players (e.g., MP3 players), cameras, game consoles, tablet devices, smart devices, wearable devices, vehicles, electric meters, gas pumps, large or small kitchen appliances, healthcare equipment, implants, sensors / actuators, displays, or any other similarly functional devices. Some of the UEs 104 may be referred to as IoT devices (e.g., parking meters, gas pumps, toasters, vehicles, heart monitors, etc.). UE 104 may also be referred to as stations, mobile stations, subscriber stations, mobile units, subscriber units, wireless units, remote units, mobile devices, wireless devices, wireless communication devices, remote devices, mobile subscriber stations, access terminals, mobile terminals, wireless terminals, remote terminals, handsets, user agents, mobile clients, clients, or some other suitable terms. In some scenarios, the term UE may also be applied to one or more supporting devices, such as in a device cluster arrangement. One or more of these devices may access the network collectively and / or individually.

[0061] Refer again Figure 1 In some aspects, the UE 104 may include a federated ML parameter reporting component 198 configured to: receive parameters of a machine learning model from a network node; calculate a gradient value relative to one of the parameters of the machine learning model, the gradient value comprising a positive gradient value or a negative gradient value; and send an analog signal indicating a magnitude of the gradient value relative to the parameter in the parameters of the machine learning model to the network node in one of a pair of REs, the pair of REs being designated for indicating the gradient value relative to the parameter in the parameters of the machine learning model. In some aspects, the federated ML parameter reporting component 198 may be configured to: receive parameters of a machine learning model from a network node; calculate a gradient value relative to one of the parameters of the machine learning model, the gradient value comprising a positive gradient value or a negative gradient value; and send an analog signal to the network node in a single resource element (RE) or skip sending the analog signal based on the gradient value relative to the parameter in the parameters of the machine learning model, the single RE being designated for indicating the positive gradient value or the negative gradient value of the gradient value relative to the parameter in the parameters of the machine learning model.

[0062] In certain aspects, the base station 102 may include a federal ML parameter processing component 199, which is configured to: output parameters of a machine learning model for multiple UEs for transmission; obtain at least one aggregated analog signal in a pair of REs designated for multiple UEs to indicate a gradient value of a parameter in the parameters of the machine learning model of the multiple UEs, the at least one aggregated analog signal obtained in a pair of REs representing multiple analog signals cumulatively obtained from multiple UEs at a network node, each of the multiple analog signals indicating a gradient value of the parameter in the parameters of the machine learning model calculated at a corresponding UE among the multiple UEs, the gradient value including a positive gradient value or a negative gradient value; and generate updated parameter information of the parameter in the parameters of the machine learning model based on the at least one aggregated analog signal obtained in a pair of REs. In some aspects, the federated ML parameter processing component 199 may be configured to: output parameters of a machine learning model for multiple UEs for transmission; obtain an aggregated analog signal in a single RE designated for multiple UEs to indicate a gradient value of a parameter in the parameters of the machine learning model for multiple UEs, the aggregated analog signal obtained in the single RE representing multiple analog signals cumulatively obtained from multiple UEs at a network node, each of the multiple analog signals indicating that the gradient value of the parameter in the parameters of the machine learning model calculated at a corresponding UE in the multiple UEs is one of a positive gradient value or a negative gradient value; and generate at least one updated parameter information of the parameter in the parameters of the machine learning model based on the aggregated analog signal obtained in the single RE.

[0063] Although the following description may focus on 5G NR, the concepts described herein may be applicable to other similar areas such as LTE, LTE-A, CDMA, GSM, and other wireless technologies.

[0064] Figure 2A FIG200 is a diagram illustrating an example of a first subframe within a 5G NR frame structure. Figure 2B is a diagram 230 illustrating an example of DL channels within a 5G NR subframe. Figure 2C 250 is a diagram illustrating an example of a second subframe within a 5G NR frame structure. Figure 2D FIG280 is a diagram illustrating an example of UL channels within a 5G NR subframe. The 5G NR frame structure may be frequency division duplex (FDD) (wherein for a particular set of subcarriers (carrier system bandwidth), subframes within that subcarrier set are dedicated to either DL or UL), or may be time division duplex (TDD) (wherein for a particular set of subcarriers (carrier system bandwidth), subframes within that subcarrier set are dedicated to both DL and UL). Figure 2A , Figure 2CIn the example provided, the 5G NR frame structure is assumed to be TDD, where subframe 4 is configured with slot format 28 (most of which are DL), where D is DL, U is UL, and F is flexible between DL / UL, and subframe 3 is configured with slot format 1 (all of which are UL). Although subframes 3 and 4 are shown as having slot formats 1 and 28, respectively, any particular subframe may be configured with any of the various available slot formats 0 to 61. Slot formats 0 and 1 are DL and UL, respectively. Other slot formats 2 to 61 include a mix of DL, UL and flexible symbols. The UE is configured with the slot format through the received slot format indicator (SFI) (dynamically configured through DL control information (DCI) or semi-statically / statically configured through radio resource control (RRC) signaling). Note that the following description also applies to the 5GNR frame structure as TDD.

[0065] FIG. 2A to FIG. 2D The frame structure is illustrated, and various aspects of the present disclosure may be applicable to other wireless communication technologies that may have different frame structures and / or different channels. A frame (10ms) may be divided into 10 subframes of the same size (1ms). Each subframe may include one or more time slots. A subframe may also include a microslot, which may include 7, 4 or 2 symbols. Each time slot may include 14 or 12 symbols, depending on whether the cyclic prefix (CP) is conventional or extended. For a conventional CP, each time slot may include 14 symbols, and for an extended CP, each time slot may include 12 symbols. The symbol on the DL may be a CP orthogonal frequency division multiplexing (OFDM) (CP-OFDM) symbol. The symbol on the UL may be a CP-OFDM symbol (for high throughput scenarios) or a discrete Fourier transform (DFT) extended OFDM (DFT-s-OFDM) symbol (also known as a single carrier frequency division multiple access (SC-FDMA) symbol) (for power-limited scenarios; limited to single-stream transmission). The number of time slots within a subframe is based on the CP and parameter set. The parameter set defines the subcarrier spacing (SCS) and effectively defines the symbol length / duration, which is equal to 1 / SCS.

[0066]

[0067]

[0068] For normal CP (14 symbols / slot), different parameter sets μ0 to 4 allow 1, 2, 4, 8, and 16 slots per subframe, respectively. For extended CP, parameter set 2 allows 4 slots per subframe. Accordingly, for normal CP and parameter set μ, there are 14 symbols / slot and 2 μ timeslots / subframe. The subcarrier spacing can be equal to 2 μ*15kHz, where μ is parameter set 0 to 4. Therefore, the subcarrier spacing for parameter set μ=0 is 15kHz, and the subcarrier spacing for parameter set μ=4 is 240kHz. The symbol length / duration is inversely related to the subcarrier spacing. FIG. 2A to FIG. 2D An example of a parameter set μ=2 with a normal CP of 14 symbols per slot and 4 slots per subframe is provided. The slot duration is 0.25 ms, the subcarrier spacing is 60 kHz, and the symbol duration is about 16.67 μs. Within a frame set, there may be one or more different bandwidth parts (BWPs) that are frequency-division multiplexed (see Figure 2B ). Each BWP may have a specific parameter set and CP (normal or extended).

[0069] A resource grid may be used to represent a frame structure. Each slot includes a resource block (RB) (also referred to as a physical RB (PRB)) extending over 12 consecutive subcarriers. The resource grid is divided into a number of resource elements (REs). The number of bits carried by each RE depends on the modulation scheme.

[0070] like Figure 2A As illustrated, some of the REs carry reference (pilot) signals (RS) for the UE. The RS may include a demodulation RS (DM-RS) (indicated as R for a particular configuration, but other DM-RS configurations are possible) and a channel state information reference signal (CSI-RS) for channel estimation at the UE. The RS may also include a beam measurement RS (BRS), a beam refinement RS (BRRS), and a phase tracking RS (PT-RS).

[0071] Figure 2BExamples of various DL channels within a subframe of a frame are illustrated. A physical downlink control channel (PDCCH) carries DCI within one or more control channel elements (CCEs) (e.g., 1, 2, 4, 8, or 16 CCEs), each CCE including six RE groups (REGs), each REG including 12 consecutive REs in an OFDM symbol of an RB. A PDCCH within a BWP may be referred to as a control resource set (CORESET). The UE is configured to monitor PDCCH candidates in a PDCCH search space (e.g., a common search space, a UE-specific search space) during a PDCCH monitoring opportunity on a CORESET, wherein the PDCCH candidates have different DCI formats and different aggregation levels. Additional BWPs may be located at higher and / or lower frequencies on the channel bandwidth. A primary synchronization signal (PSS) may be within symbol 2 of a particular subframe of a frame. The PSS is used by UE 104 to determine subframe / symbol timing and physical layer identification. A secondary synchronization signal (SSS) may be within symbol 4 of a particular subframe of a frame. The SSS is used by the UE to determine the physical layer cell identity group number and the radio frame timing. Based on the physical layer identity and the physical layer cell identity group number, the UE can determine the physical cell identifier (PCI). Based on the PCI, the UE can determine the location of the DM-RS. The physical broadcast channel (PBCH) carrying the master information block (MIB) can be logically grouped with the PSS and SSS to form a synchronization signal (SS) / PBCH block (also referred to as an SS block (SSB)). The MIB provides the number of RBs in the system bandwidth and the system frame number (SFN). The physical downlink shared channel (PDSCH) carries user data, broadcast system information not sent via the PBCH (such as a system information block (SIB)), and paging messages.

[0072] like Figure 2C As illustrated, some of the REs carry DM-RS (indicated as R for a particular configuration, but other DM-RS configurations are possible) for channel estimation at the base station. The UE may send a DM-RS for a physical uplink control channel (PUCCH) and a DM-RS for a physical uplink shared channel (PUSCH). The PUSCH DM-RS may be sent in the first or first two symbols of the PUSCH. Depending on whether a short PUCCH or a long PUCCH is sent and depending on the specific PUCCH format used, the PUCCH DM-RS may be sent in different configurations. The UE may send a sounding reference signal (SRS). The SRS may be sent in the last symbol of a subframe. The SRS may have a comb structure, and the UE may send the SRS on one of the comb teeth. The SRS may be used by the base station for channel quality estimation to achieve frequency-dependent scheduling of the UL.

[0073] Figure 2DExamples of various UL channels within a subframe of a frame are illustrated. The PUCCH may be located at a position as indicated in one configuration. The PUCCH carries uplink control information (UCI), such as scheduling requests, channel quality indicators (CQI), precoding matrix indicators (PMI), rank indicators (RI), and hybrid automatic repeat request (HARQ) acknowledgement (ACK) (HARQ-ACK) feedback (i.e., one or more HARQ ACK bits indicating one or more ACKs and / or negative ACKs (NACKs)). The PUSCH carries data and may additionally be used to carry a buffer status report (BSR), a power headroom report (PHR), and / or UCI.

[0074] Figure 3 375. The Controller / Processor 375 is a block diagram of a base station 310 in an access network communicating with a UE 350. In the DL, Internet Protocol (IP) packets may be provided to a controller / processor 375. The controller / processor 375 implements layer 3 and layer 2 functionality. Layer 3 includes a radio resource control (RRC) layer, and layer 2 includes a service data adaptation protocol (SDAP) layer, a packet data convergence protocol (PDCP) layer, a radio link control (RLC) layer, and a medium access control (MAC) layer. The controller / processor 375 provides RRC layer functionality associated with broadcasting of system information (e.g., MIB, SIB), RRC connection control (e.g., RRC connection paging, RRC connection establishment, RRC connection modification, and RRC connection release), inter-radio access technology (RAT) mobility, and measurement configuration for UE measurement reporting; PDCP layer functionality associated with header compression / decompression, security (encryption, decryption, integrity protection, integrity verification), and handover support functions; RLC layer functionality associated with delivery of upper layer packet data units (PDUs), error correction through ARQ, concatenation, segmentation and reassembly of RLC service data units (SDUs), re-segmentation of RLC data PDUs, and reordering of RLC data PDUs; and MAC layer functionality associated with mapping between logical channels and transport channels, multiplexing of MAC SDUs onto transport blocks (TBs), demultiplexing of MAC SDUs from TBs, scheduling information reporting, error correction through HARQ, priority handling, and logical channel prioritization.

[0075] The transmit (TX) processor 316 and receive (RX) processor 370 implement layer 1 functionality associated with various signal processing functions. Layer 1 (which includes the physical (PHY) layer) may include error detection on the transmission channel, forward error correction (FEC) coding / decoding of the transmission channel, interleaving, rate matching, mapping to the physical channel, modulation / demodulation of the physical channel, and MIMO antenna processing. The TX processor 316 processes the mapping to the signal constellation based on various modulation schemes (e.g., binary phase shift keying (BPSK), quadrature phase shift keying (QPSK), M-order phase shift keying (M-PSK), M-order quadrature amplitude modulation (M-QAM)). The coded and modulated symbols can then be split into parallel streams. Each stream can then be mapped to an OFDM subcarrier, multiplexed with a reference signal (e.g., a pilot) in the time and / or frequency domain, and then each stream is combined together using an inverse fast Fourier transform (IFFT) to produce a physical channel carrying a time domain OFDM symbol stream. The OFDM stream is spatially pre-coded to generate multiple spatial streams. Channel estimates from a channel estimator 374 may be used to determine coding and modulation schemes, as well as for spatial processing. Channel estimates may be derived based on reference signals and / or channel condition feedback sent by the UE 350. Each spatial stream may then be provided to a different antenna 320 via a separate transmitter 318Tx. Each transmitter 318Tx may modulate a radio frequency (RF) carrier for transmission using a corresponding spatial stream.

[0076] At the UE 350, each receiver 354Rx receives a signal through its corresponding antenna 352. Each receiver 354Rx recovers the information modulated onto the RF carrier and provides the information to a receive (RX) processor 356. The TX processor 368 and the RX processor 356 implement layer 1 functionality associated with various signal processing functions. The RX processor 356 can perform spatial processing on the information to recover any spatial stream destined for the UE 350. If multiple spatial streams are destined for the UE 350, they can be combined into a single OFDM symbol stream by the RX processor 356. The RX processor 356 then converts the OFDM symbol stream from the time domain to the frequency domain using a fast Fourier transform (FFT). The frequency domain signal includes a separate OFDM symbol stream for each subcarrier of the OFDM signal. The symbol on each subcarrier and the reference signal are recovered and demodulated by determining the signal constellation point most likely to be sent by the base station 310. These soft decisions can be based on the channel estimation value calculated by the channel estimator 358. The soft decisions are then decoded and deinterleaved to recover the data and control signals originally sent on the physical channel by base station 310. The data and control signals are then provided to controller / processor 359, which implements layer 3 and layer 2 functionality.

[0077] The controller / processor 359 may be associated with a memory 360 that stores program codes and data. The memory 360 may be referred to as a computer readable medium. In the UL, the controller / processor 359 provides demultiplexing between transport channels and logical channels, packet reassembly, decryption, header decompression, and control signal processing to recover IP packets. The controller / processor 359 is also responsible for error detection using an ACK protocol and / or a NACK protocol to support HARQ operations.

[0078] Similar to the functionality described in conjunction with DL transmissions performed by the base station 310, the controller / processor 359 provides RRC layer functionality associated with system information (e.g., MIB, SIB) acquisition, RRC connection, and measurement reporting; PDCP layer functionality associated with header compression / decompression and security (encryption, decryption, integrity protection, integrity verification); RLC layer functionality associated with delivery of upper layer PDUs, error correction through ARQ, concatenation, segmentation and reassembly of RLC SDUs, resegmentation of RLC data PDUs, and reordering of RLC data PDUs; and MAC layer functionality associated with mapping between logical channels and transport channels, multiplexing of MAC SDUs onto TBs, demultiplexing of MAC SDUs from TBs, scheduling information reporting, error correction through HARQ, priority handling, and logical channel prioritization.

[0079] The TX processor 368 may use channel estimates derived by the channel estimator 358 from a reference signal or feedback sent by the base station 310 to select appropriate coding and modulation schemes and to facilitate spatial processing. The spatial streams generated by the TX processor 368 may be provided to different antennas 352 via separate transmitters 354Tx. Each transmitter 354Tx may modulate an RF carrier with a corresponding spatial stream for transmission.

[0080] UL transmissions are processed at the base station 310 in a manner similar to that described in conjunction with the receiver functionality at the UE 350. Each receiver 318Rx receives a signal through its respective antenna 320. Each receiver 318Rx recovers information modulated onto an RF carrier and provides the information to the RX processor 370.

[0081] The controller / processor 375 may be associated with a memory 376 that stores program codes and data. The memory 376 may be referred to as a computer readable medium. In the UL, the controller / processor 375 provides demultiplexing between transport channels and logical channels, packet reassembly, decryption, header decompression, control signal processing to recover IP packets. The controller / processor 375 is also responsible for error detection using ACK protocol and / or NACK protocol to support HARQ operations.

[0082] At least one of the TX processor 368, the RX processor 356, and the controller / processor 359 may be configured to perform operations related to Figure 1 At least one of the TX processor 316, the RX processor 370, and the controller / processor 375 may be configured to perform operations related to Figure 1 The federated ML parameter processing component 199 combines various aspects.

[0083] Figure 4 400 is an example of an AI / ML algorithm for a method of wireless communication. The AI / ML algorithm 400 may include various functions, including a data collection function 402, a model training function 404, a model inference function 406, and participants 408.

[0084] The data collection function 402 may be a function that provides input data to the model training function 404 and the model inference function 406. The data collection function 402 may include any form of data preparation, and it may not be specific to the specific implementation of the AI / ML algorithm (e.g., data preprocessing and cleaning, formatting, and transformation). Examples of input data may include, but are not limited to, measurements from network entities including UEs or network nodes, feedback from participants 408, and output from another AI / ML model. The data collection function 402 may include training data and inference data, wherein the training data refers to data to be transmitted as input to the model training function 404, and the inference data refers to data to be transmitted as input to the model inference function 406.

[0085] The model training function 404 may be a function that performs ML model training, validation, and testing, which may generate model performance metrics as part of the model testing procedure. The model training function 404 may also be responsible for performing data preparation (e.g., data preprocessing and cleaning, formatting, and transformation) based on the training data delivered or received from the data collection function 402. The model training function 404 may deploy or update the trained, validated, and tested AI / ML model to the model inference function 406, and receive model performance feedback from the model inference function 406.

[0086] The model inference function 406 may be a function that provides a model inference output (e.g., a prediction or decision). The model inference function 406 may also perform data preparation (e.g., data preprocessing and cleaning, formatting, and transformation) based on the inferred data delivered from the data collection function 402. The output of the model inference function 406 may include the inferred output of the AI / ML model generated by the model inference function 406. The details of the inferred output may be use case specific.

[0087] Model performance feedback may refer to information derived from model inference functionality 406 that may be applicable to improving the AI / ML model trained in model training functionality 404. Feedback from participants 408 or other network entities (via data collection functionality 402) may be implemented for model inference functionality 406 to create model performance feedback.

[0088] Participants 408 may be functions that receive output from model inference function 406 and trigger or perform corresponding actions. Participants 408 may trigger actions for network entities including other network entities or themselves. Participants 408 may also provide feedback information, and model training function 404 or model inference function 406 may derive training data or inference data or performance feedback. This feedback may be sent back to data collection function 402.

[0089] The UE and / or network entities (centralized units and / or distributed units) may use machine learning algorithms, deep learning algorithms, neural networks, reinforcement learning, regression, boosting or advanced signal processing methods for various aspects of wireless communications, such as with a base station, a TRP, another UE, etc.

[0090] In some aspects described herein, a coding device (e.g., a UE) may train one or more neural networks to learn the dependence of the measured quality on various parameters. Examples of machine learning models or neural networks that may be included in a UE and / or a network entity include, among others: an artificial neural network (ANN); decision tree learning; a convolutional neural network (CNN); a deep learning architecture in which the output of a first layer of neurons becomes the input of a second layer of neurons, etc.; a support vector machine (SVM), for example, including a separating hyperplane (e.g., a decision boundary) for classifying data; regression analysis; a Bayesian network; a genetic algorithm; a deep convolutional network (DCN) configured with additional pooling and normalization layers; and a deep belief network (DBN).

[0091] A machine learning model, such as an artificial neural network (ANN), may include a set of interconnected artificial neurons (e.g., a neuron model) and may be a computing device or may represent a method to be performed by a computing device. The connections of the neuron model may be modeled as weights. The machine learning model may provide predictive models, adaptive control, and other applications by training via a data set. The model may be adaptive based on external or internal information processed by the machine learning model. Machine learning may provide nonlinear statistical data models or decision making, and may model complex relationships between input data and output information.

[0092] The machine learning model may include multiple layers and / or operations, which may be formed by cascading one or more of the operations cited. Examples of operations that may be involved include: extraction of various features of data, convolution operations, fully connected operations that may be activated or deactivated, compression, decompression, quantization, flattening, etc. As used herein, the "layer" of the machine learning model can be used to represent operations on input data. For example, convolution layers and / or fully connected layers, etc. can be used to refer to associated operations on data input to the layer. Convolution A×B operations refer to operations that convert multiple input features A into multiple output features B. "Kernel size" may refer to the number of neighboring coefficients combined in a dimension. As used herein, "weights" can be used to represent one or more coefficients used in operations for combining individual rows and / or columns of input data in each layer. For example, a fully connected layer operation may have an output y, which is determined at least in part based on the sum of the product of the input matrix x and the weight A (which may be a matrix) and the bias value B (which may be a matrix). The term "weight" can generally be used herein to refer to both weights and bias values. Weights and biases are examples of parameters of a trained machine learning model.Different layers of a machine learning model can be trained separately.

[0093] Machine learning models may include various connectivity patterns, for example, including any of feedforward networks, hierarchies, recursive architectures, feedback connections, and the like. The connections between layers of a neural network may be fully connected or locally connected. In a fully connected network, a neuron in a first layer may communicate its output to each neuron in a second layer, and each neuron in a second layer may receive input from each neuron in the first layer. In a locally connected network, a neuron in a first layer may be connected to a limited number of neurons in a second layer. In some aspects, a convolutional network may be locally connected and configured with a shared connection strength associated with the input of each neuron in the second layer. A locally connected layer of a network may be configured so that each neuron in the layer has the same or similar connectivity pattern, but with different connection strengths.

[0094] A machine learning model or neural network may be trained. For example, a machine learning model may be trained based on supervised learning. During training, the machine learning model may be presented with inputs that the model uses to calculate to produce an output. The actual output may be compared to the target output, and the difference may be used to adjust the parameters of the machine learning model (such as weights and biases) to provide an output closer to the target output. Before training, the output may be incorrect or less accurate, and the error or difference between the actual output and the target output may be calculated. Then, the weights of the machine learning model may be adjusted so that the output is more closely aligned with the target. In order to adjust the weights, the learning algorithm may calculate a gradient vector for the weights. The gradient may indicate the amount by which the error will increase or decrease if the weights are slightly adjusted. At the top layer, the gradient may directly correspond to the value of the weights connecting the activated neurons in the penultimate layer and the neurons in the output layer. In the lower layers, the gradient may depend on the value of the weights and the calculated error gradients of the higher layers. Then, the weights may be adjusted to reduce the error or move the output closer to the target. This way of adjusting the weights may be referred to as back propagation through a neural network. The process may continue until the achievable error rate stops decreasing, or until the error rate has reached a target level.

[0095] These machine learning models may include computational complexity and a large number of processors for training machine learning models. The output of one node is connected as an input to another node. The connection between nodes may be referred to as an edge, and weights may be applied to the connection / edge to adjust the output from one node, which is used as an input to another node. Nodes may apply thresholds to determine whether or when to provide output to a connected node. The output of each node may be calculated as a nonlinear function of the sum of the inputs to the node. A neural network may include any number of nodes and any type of connection between the nodes. A neural network may include one or more hidden nodes. Nodes may be aggregated into layers, and different layers of a neural network may perform different kinds of transformations on the input. A signal may travel from an input at the first layer through multiple layers of a neural network to an output at the last layer of the neural network, and may traverse these layers multiple times.

[0096] In some aspects, multiple UEs may include a machine learning model with global parameters, and various implementations for training the global parameters of the machine learning model for the multiple UEs may be provided. In one aspect, the multiple UEs may send training data to a network node, and the network node may train, adjust the global parameters based on the training data received from the multiple UEs, and send updated global parameters to the multiple UEs. Here, communicating the training data from the multiple UEs to the network node may result in a large amount of traffic.

[0097] In another aspect, a federated learning implementation may be provided for training global parameters of a machine learning model for multiple UEs. Federated learning may refer to a method of training global parameters of a machine learning model through wireless communication, wherein a client (e.g., UE) may train a local machine learning model and provide the trained local parameters to a central server (e.g., a network node), and the network node may update the global parameters of the machine learning model and distribute the updated global parameters to the clients (e.g., UEs). The federated learning method may achieve efficient learning by utilizing training data from many different clients (e.g., all available UEs) and without the need to exchange actual local training data (e.g., to communicate model-related information).

[0098] An over-the-air (OTA) aggregation scheme may be implemented to reduce the overhead specified for sending local model information over wireless resources. That is, based on OTA aggregation, data of local model information from multiple UEs may be sent in an aggregated manner, and a network node may receive the aggregated data to determine the sum of local model information from multiple UEs. OTA aggregation may generally be associated with tight synchronization (especially in-phase synchronization), which may be challenging. On the other hand, a non-coherent OTA aggregation scheme may be implemented to reduce the tight phase synchronization specification and perform a conventional OTA scheme with perfect phase synchronization while improving resource utilization.

[0099] Federated learning can provide an effective means of training global machine learning models in a semi-distributed manner. In federated learning, multiple UEs (e.g., multiple clients) can adopt the same machine learning model structure for a task of interest, and multiple UEs can locally train their model parameters based on their observations (e.g., local training data). Each of the multiple clients can periodically transmit information about updated (e.g., locally trained) model parameters to a server (e.g., a network node), and the server can compile the information of the updated model parameters received from all multiple clients to create global (e.g., aggregated) model parameters, which can then be forwarded to multiple clients to be applied to local machine learning models.

[0100] In one example, stochastic gradient descent (SGD) model training may be implemented to compile information of updated model parameters received from all multiple clients, and the updated global parameters generated by the server at iteration i (e.g., at a specific time point or epoch) may be represented as where θ i A vector representing the global model parameters at iteration i, It can represent the gradient vector value calculated by the nth UE at iteration i, N UEsmay represent the number of UEs participating in the i-th round (or iteration), and n i It can represent the learning rate.

[0101] Here, for simplicity, the above formula may assume that each UE performs local updates using a single training data, but aspects of the present disclosure are not limited thereto. In general, local training of a machine learning model may be performed using multiple training data, and the number of training data may be different between multiple UEs, and the weighted sum of the gradient values ​​may be configured to calculate a vector of global model parameters.

[0102] For example, a typical machine learning model may have a large number of machine learning parameters (e.g., tens or hundreds of parameters out of thousands of parameters). In a wireless communication environment of multiple UEs facilitating a federated learning training session, the multiple UEs may be configured to send information about the gradients of a large number of machine learning parameters to a server. Multiple UEs individually sending information about the gradients of local parameters may result in increased network overhead.

[0103] In one aspect, SGD model training can be specified as based on the gradient vector value for all n The updated global parameters are generated by summing the values ​​of the gradient vectors for all n. Therefore, the concept of OTA aggregation (also called OTA calculation) can be used to convert the gradient vector value Sent to the server. In a specific implementation of OTA aggregation, the server may reserve at least one RE for conveying each machine learning model parameter, and each UE may send the gradient of each model parameter on the corresponding at least one RE using an analog pulse-amplitude modulation (PAM) signal (e.g., analog modulation). For example, a PAM signal may be sent in the in-phase and quadrature dimensions, and in the case where the symbol sent on each RE is an analog complex value, the RE may accommodate two model parameters.

[0104] The server may monitor each RE and receive a superposition result (e.g., a summation result) of the gradient values ​​of the corresponding gradient machine learning parameters from all available (or participating) UEs, which is information that the server may consider updating the corresponding model parameters based on SGD model training. Effectively, OTA aggregation may enable federated learning to scale relative to the number of UEs participating in the learning, thereby achieving a reduction in communication overhead by a factor proportional to the number of UEs participating in the federated learning. Considering that the machine learning model includes a large number of parameters (e.g., tens or hundreds of parameters out of thousands of parameters), the reduction in communication overhead may be significant.

[0105] Wireless communication can be performed in the presence of channel imperfections such as multipath, frequency and time / clock offsets between transmission and reception, etc. In one example, instead of receiving Servers using OTA aggregation can observe in may represent the channel impairment (including path loss effects) experienced by the nth UE and may be a complex value, may represent the transmit power of the UE (e.g., when the gradient has a normalized mean power), and May represent a complex rotation that may be applied by the UE to compensate for phase distortion due to channel and timing / frequency offsets. Here, the example provides experiencing the same channel conditions and using the same transmit power on all REs, but the presently disclosed aspects are not limited thereto, and different REs may experience different channel conditions and the transmit power used by the UE may be different.

[0106] To accommodate OTA aggregation implementations, channel conditions and UE implementations can be configured to achieve Where C may be a constant (e.g., one (1)) for all UEs (e.g., identified by index n). Thus, the wireless network may be configured with tight synchronization including power control to compensate for path loss and phase synchronization to compensate for phase misalignment due to multipath propagation, carrier frequency offset, and timing mismatch between each UE and the server.

[0107] In addition to configuring wireless networks with tight synchronization for OTA aggregation rounds, clock / oscillator drift between the time when the synchronization procedure is completed and the time when the OTA aggregation transmission is performed can lead to significant uncertainty in phase synchronization. BB In the case of (t) (e.g., a waveform carrying two gradient values ​​on the in-phase component and the quadrature component) and a timing offset t0>0 relative to the server time reference, the signal observed by the server may be generated as follows: Even with a small time offset (e.g., 0.17 ns for fc = 5.9 GHz), the phase rotation φ0 = 2πf c t0 may also become equal to 2π (total phase uncertainty). According to the current example, sample level timing accuracy may be configured to achieve phase synchronicity, which may be very difficult to achieve and maintain, especially in OFDM transmission schemes including cyclic prefix level synchronization, which is coarser than sample level synchronization. Therefore, the OTA aggregation scheme may be configured to be insensitive or robust to phase uncertainty. The OTA aggregation scheme that is insensitive or robust to phase uncertainty may relax the phase synchronization configuration of OTA and can provide federated learning with OTA aggregation implementation for wireless communication.

[0108] Figure 5 is a first example of a machine learning parameter indicating a method of wireless communication. The first example indicating a machine learning parameter may be a non-coherent OTA aggregation including a signed SGD model for federated training of machine learning parameters. In the signed SGD model, UEs may provide signs (e.g., + or -) of their local gradients to a server (and not provide gradient magnitude information) instead of reporting their local gradients, and the server may perform a majority voting binary decision (e.g., the gradient value sent to the server may be set to +1 or -1, depending on whether more UEs report positive gradient signs or negative gradient signs, respectively). Various aspects of the present disclosure provide that a gradient value may be sent as a PAM signal having a +1 value or a -1 value, but the gradient value may be set to have a different constant value (e.g., +p or -p, where p is a real number). The signed SGD model may be an effective low-complexity or low-overhead implementation of an SGD model for machine learning training.

[0109] Compared to allocating one (1) RE for every 2 gradients sent using I / Q signals, a specific implementation of the first example of non-coherent OTA aggregation including a signed SGD model for federated training of machine learning parameters may include reserving two REs (e.g., a pair of REs) for each gradient element. Here, a pair of REs may include a first RE that is a positive RE and a second RE that is a negative RE. That is, a server may (e.g., through a network node) allocate a pair of REs for multiple UEs to send gradient values ​​using an analog signal (e.g., a PAM signal), the pair of REs including a positive RE dedicated to multiple UEs to convey an analog signal corresponding to a positive gradient value and a negative RE dedicated to multiple UEs to convey an analog signal corresponding to a negative gradient value.

[0110] For a given gradient value, each UE may send a constant value corresponding to the sign of the gradient value (e.g., +p, where p is a real number) on one RE and may not send any signal on another RE. For example, for a positive gradient value, the UE may send an analog signal representing a value of +1 on the positive RE of the RE pair, and for a negative gradient value, the UE may send an analog signal representing a value of +1 on the negative RE of the RE pair.

[0111] Figure 55 shows how the UE may map four (4) locally calculated gradient values ​​500 to transmitted values ​​on corresponding sign REs 510 and 520. A server may allocate a pair of REs 510 and 520 to the UE (e.g., via a network node), and the pair of REs 510 and 520 may include a positive sign RE 510 and a negative sign RE 520. For example, for a positive gradient value (e.g., 1.3 or 0.5), the UE may transmit an analog signal representing a value of +1 in the positive sign RE 510, and for a negative gradient value (e.g., -0.4 or -1.2), the UE may transmit an analog signal representing a value of +1 in the negative sign RE 520.

[0112] Assuming ideal transmit power control, the first aggregate analog signal received at the server in the positive RE can be expressed as Where N + It can represent the number of UEs with positive gradient values, can represent the magnitude of the positive gradient value, and may represent an arbitrary phase experienced by a UE with index n due to a time-phase-frequency error; and the second aggregate analog signal received at the server in the negative RE may be represented as Where N - It can represent the number of UEs with negative gradient values, can represent the magnitude of the positive gradient value, and It can be represented by the time-phase-frequency error caused by the index n - Any phase experienced by the UE.

[0113] The server may determine the aggregate gradient value based on the first aggregate analog signal received in the positive RE and the second aggregate analog signal received in the negative RE. In one aspect, the server may compare the first magnitude of the first aggregate analog signal received in the positive RE with the second magnitude of the second aggregate analog signal received in the negative RE. The server may compare |y + | 2 with |y - | 2 Compare and based on |y + | 2 with |y - | 2 In one example, based on determining that the first magnitude of the first aggregate analog signal is greater than the second magnitude of the second aggregate analog signal (eg, |y + | 2 >|y - | 2), the server may declare that the aggregate gradient value is +1. In another example, based on determining that the first magnitude of the first aggregate analog signal is less than or equal to the second magnitude of the second aggregate analog signal (e.g., |y + | 2 ≤|y - | 2 ), the server may declare that the aggregate gradient value is -1. The server may update the parameters based on the aggregate gradient value and broadcast (or send) the updated parameters to multiple UEs.

[0114] According to Figure 5 In the illustrated first example of non-coherent OTA aggregation, the server may train global parameters with relaxed phase synchronization. The aggregate gradient values ​​calculated based on the first example of non-coherent OTA aggregation may be correlated even under total phase uncertainty (e.g., the phase is randomly distributed between 0 and 2π with enough clients / UEs). The non-coherent OTA aggregation scheme may also relax the specifications on power control and does not consider the number of UEs participating in each round. Under the first example of non-coherent OTA aggregation, multiple UEs may send unit magnitude signals on corresponding token REs instead of sending actual gradient values, and therefore, some information may be lost in each training round, which may result in slower convergence.

[0115] Figure 6 is a second example of a machine learning parameter indicating a method of wireless communication. The second example indicating the machine learning parameter may be a non-coherent OTA aggregation including a modified sign SGD model for federated training of the machine learning parameter. The second example of the non-coherent OTA aggregation including the modified sign SGD model for federated training of the machine learning parameter may follow substantially the same scheme as the first example of the non-coherent OTA aggregation, except that, instead of transmitting a PAM signal representing a constant value (e.g., +1), the UE may send an absolute magnitude of the gradient value on a corresponding sign RE (e.g., a positive sign RE or a negative sign RE).

[0116] A specific implementation of the second example of non-coherent OTA aggregation may include two REs (e.g., a pair of REs) reserved for each gradient element. Here, a pair of REs may include a first RE that is a positive RE and a second RE that is a negative RE. That is, a server may (e.g., through a network node) allocate a pair of REs to multiple UEs to send gradient values ​​using an analog signal (e.g., a PAM signal), the pair of REs including a positive RE dedicated to multiple UEs to convey an analog signal corresponding to a positive gradient value and a negative RE dedicated to multiple UEs to convey an analog signal corresponding to a negative gradient value.

[0117] For a given gradient value, each UE may send a value corresponding to the absolute magnitude of the gradient on one RE corresponding to the sign of the gradient value, and may not send any signal on the other RE. For example, for a positive gradient value, the UE may send an analog signal representing the absolute magnitude of the gradient value on the positive RE of the RE pair, and for a negative gradient value, the UE may send an analog signal representing the absolute magnitude of the gradient value on the negative RE of the RE pair.

[0118] Figure 6 6 shows how the UE may map four (4) locally calculated gradient values ​​600 to values ​​transmitted on corresponding sign REs 610 and 620. A server may allocate a pair of REs 610 and 620 to the UE (e.g., via a network node), and the pair of REs 610 and 620 may include a positive sign RE 610 and a negative sign RE 620. For example, for a given gradient, the UE may observe a positive value of +1.3 or a positive value of +0.5, and the UE may transmit a value of +1.3 or a value of +0.5 on the corresponding positive sign RE 610. For a given gradient, the UE may observe a value of -0.4 or a value of -1.2, and the UE may transmit a value of +0.4 or a value of +1.2 on the corresponding negative sign RE 620.

[0119] Assuming ideal transmit power control, the first aggregate analog signal received at the server in the positive RE can be expressed as Where N + It can represent the number of UEs with positive gradient values, can represent the magnitude of the positive gradient value, and It can be represented by the time-phase-frequency error caused by the index n + The second aggregate analog signal received at the server in the negative RE can be expressed as Where N - It can represent the number of UEs with negative gradient values, can represent the magnitude of the positive gradient value, and It can be represented by the time-phase-frequency error caused by the index n - Any phase experienced by the UE.

[0120] The server may determine the aggregate gradient value based on the first aggregate analog signal received in the positive RE and the second aggregate analog signal received in the negative RE. In one aspect, the server may compare the first magnitude of the first aggregate analog signal received in the positive RE with the second magnitude of the second aggregate analog signal received in the negative RE. The server may compare |y + | 2 with |y - | 2 Compare and based on |y+ | 2 with |y - | 2 In one example, based on determining that the first magnitude of the first aggregate analog signal is greater than the second magnitude of the second aggregate analog signal (eg, |y + | 2 >|y-| 2 ), the server may declare that the aggregate gradient value is +1. In another example, based on determining that the first magnitude of the first aggregate analog signal is less than or equal to the second magnitude of the second aggregate analog signal (e.g., |y + | 2 ≤|y-| 2 ), the server may declare that the aggregate gradient value is -1. The server may update the parameters based on the aggregate gradient value and broadcast (or send) the updated parameters to multiple UEs.

[0121] According to Figure 6 In the illustrated second example of non-coherent OTA aggregation, the server may utilize relaxed phase synchronization to train global parameters. In addition, the energy of the aggregated analog signal received in a pair of REs (e.g., positive and negative) may be proportional to the number of UEs that calculated the positive or negative gradient, while adding a confidence indication for the gradient sign. In one example, a first UE that observes a positive gradient value of +1.4 may have more confidence than a second UE that observes a positive gradient value of +0.1. A first UE that sends an analog signal corresponding to a gradient value of +1.4 may have a greater impact on the aggregated gradient value than a second UE that sends an analog signal corresponding to a gradient value of +0.1. In another example, half of the multiple UEs may calculate a gradient value of +1.6, and half of the multiple UEs may calculate a gradient value of -0.001. Based on the Figure 5 In the illustrated first example of non-coherent OTA aggregation, the first aggregated analog signal received in the positive RE and the second aggregated analog signal received in the negative RE (in the usual case) may have the same energy, and the server may have an equal probability to decide whether the aggregate gradient value is +1 or -1, even though +1 would be a more reasonable result. Figure 6 In the illustrated second example of non-coherent OTA aggregation, the first aggregated analog signal received in the positive RE may have greater energy (in general), and therefore, the server may decide that the aggregation gradient value is +1.

[0122] Figure 7is a third example of a machine learning parameter indicating a method of wireless communication. The third example of indicating a machine learning parameter may be a non-coherent OTA aggregation including a modified signed SGD model for federated training of the machine learning parameter, including a modified signed SGD model. A specific implementation of the third example of a non-coherent OTA aggregation including a modified signed SGD model for federated training of the machine learning parameter may include: the UE performs a switch transmission of each gradient value. From the perspective of the UE, as Figure 5 The first example of non-coherent OTA aggregation illustrated may be a binary frequency shift keying (FSK) transmission per gradient, which provides for transmitting binary data at a frequency (eg, RE) corresponding to a symbol mark. Figure 7 A third example of non-coherent OTA aggregation illustrated may be on-off binary signaling per gradient.

[0123] The third example of non-coherent OTA aggregation may follow substantially the same scheme as the first example of non-coherent OTA aggregation, except that each gradient is associated with a single positive RE (e.g., no negative REs are configured). The UE may observe positive gradient values, and the UE may send an analog signal representing a constant value (e.g., a value of +1) on the positive RE, and the UE may observe negative gradient values, and the UE may defer sending a signal for that particular gradient value (e.g., a value of 0).

[0124] Figure 7 7 shows how the UE may map four (4) locally calculated gradient values ​​700 to values ​​sent on positive REs 710. A server may (e.g., via a network node) assign positive REs 710. For example, for positive gradient values ​​(e.g., 1.3 or 0.5), the UE may send an analog signal representing a value of +1 in the positive REs 710 to indicate that the corresponding gradient value has a positive sign, and for negative gradient values ​​(e.g., -0.4 or -1.2), the UE may defer sending an analog signal in the positive REs 710 to indicate that the corresponding gradient value has a negative sign.

[0125] Assuming ideal transmit power control, the first aggregate analog signal received at the server in the positive RE can be expressed as Where N + It can represent the number of UEs with positive gradient values, can represent the magnitude of the positive gradient value, and may represent any phase experienced by the UE due to time-phase-frequency error.

[0126] The server may determine the aggregate gradient value based on the first aggregate analog signal received in the positive RE and the threshold. In one aspect, the server may compare the first magnitude of the first aggregate analog signal received in the positive RE with the threshold. The server may+ | 2 is compared with the threshold and based on |y + | 2 The aggregate gradient value is determined based on the comparison result with the threshold. In one example, based on determining that the first magnitude of the first aggregate analog signal is greater than the threshold (eg, |y + | 2 >threshold), the server may declare that the aggregate gradient value is +1. In another example, based on determining that the first magnitude of the first aggregate analog signal is less than or equal to the second magnitude of the second aggregate analog signal (e.g., |y + | 2 ≤threshold), the server may declare that the aggregate gradient value is -1. The threshold may be determined or configured by the server. The server may update the parameters based on the aggregate gradient value and broadcast (or send) the updated parameters to multiple UEs.

[0127] According to Figure 7 In the third example of non-coherent OTA aggregation illustrated, the server can use relaxed phase synchronization to train global parameters. Figure 5 Compared to the illustrated first example of non-coherent OTA aggregation, the third example of non-coherent OTA aggregation may configure a single RE per gradient instead of two (2) REs, which is half of the resources configured for the first example of non-coherent OTA aggregation.

[0128] Figure 8 800 is a call flow diagram of a method of wireless communication. The call flow diagram 800 may include a first UE 802, a second UE 803, and a network node 804 for non-coherent OTA aggregation of federated training of machine learning parameters, the non-coherent OTA aggregation including a modified token SGD model. The network node 804 may configure at least one RE for multiple UEs including the first UE 802, the at least one RE being designated for sending multiple analog signals to indicate a gradient value relative to a parameter to the network node 804. The network node 804 may obtain at least one aggregated analog signal representing the multiple analog signals, generate an aggregated gradient of the parameter based on the at least one aggregated analog signal, and update the parameter based on the aggregated gradient. The network node 804 may send information associated with an updated parameter of the machine learning model to multiple UEs including the first UE 802 and the second UE 803. The multiple UEs including the first UE 802, the second UE 803, and the network node 804 may update the parameter of the machine learning model based on the updated parameter information.

[0129] At 806, the network node 804 may send parameters of the machine learning model for multiple UEs including the first UE 802 and the second UE 803. The first UE 802 may receive the parameters of the machine learning model from the network node 804. Here, the parameters of the machine learning model may be global parameters of the machine learning model, which may be shared by multiple UEs including the first UE 802 and the second UE 803.

[0130] At 808, the network node 804 may send an indication of at least one RE to a plurality of UEs including the first UE 802 and the second UE 803 to configure the first UE 802 and the second UE 803 to indicate a gradient value of the parameter in the parameters of the machine learning model. The first UE 802 may receive an indication of at least one RE from the network node 804, the at least one RE being configured for the plurality of UEs including the first UE 802 and the second UE 803 to indicate a gradient value of the parameter in the parameters of the machine learning model.

[0131] In a first aspect, the indication may configure a pair of REs for a plurality of UEs including a first UE 802 and a second UE 803. The pair of REs designated for indicating parameters of a machine learning model may include: a first RE designated for indicating a positive gradient value; and a second RE designated for indicating a negative gradient value.

[0132] In a second aspect, the indication may configure a single RE for multiple UEs including the first UE 802 and the second UE 803. Here, a single RE may be designated to indicate a positive gradient value or a negative gradient value of a gradient value of a parameter in a machine learning model relative to the parameter.

[0133] At 810, the first UE 802 may calculate a gradient value relative to one of the parameters of the machine learning model. Here, the gradient value may include a positive gradient value or a negative gradient value. The first UE 802 and the second UE 803 may respectively perform training on the parameter among the parameters received at 806, and calculate the gradient value of the parameter based on the training of the parameter. The first UE 802 and the second UE 803 may use different training data sets, so the first UE 802 and the second UE 803 may calculate different gradient values ​​based on the training of the parameter.

[0134] At 814, the first UE 802 may send an analog signal in at least one RE indicating a gradient value of the parameter in the machine learning model to the network node 804. The analog signal is sent in at least one RE configured based on the indication of at least one RE received from the network node 804 at 808.

[0135] In a first aspect, the first UE 802 may send an analog signal in one of a pair of REs to the network node 804 indicating a magnitude of a gradient value relative to the parameter in the parameters of the machine learning model. A pair of REs may be designated to indicate a gradient value relative to the parameter in the parameters of the machine learning model as configured by an indication received from the network node 804 at 808. The pair of REs designated to indicate the parameters of the machine learning model may include: a first RE, which is designated to indicate a positive gradient value; and a second RE, which is designated to indicate a negative gradient value. The first UE 802 may send the analog signal sent in the first RE or the second RE based on the gradient value. For example, the first UE 802 may observe a positive value of +1.3 or a positive value of +0.5 as a given gradient of the parameter, and the first UE 802 may send an analog signal representing a value of +1.3 or a value of +0.5 on the first RE. The first UE 802 may observe a negative value of -0.4 or a negative value of -1.2 as a given gradient of the parameter, and the first UE 802 may send a simulated signal on the second RE representing a value of +0.4 or a value of +1.2.

[0136] In a second aspect, the first UE 802 may send an analog signal to the network node 804 in a single RE or skip sending the analog signal based on the sign of the gradient value relative to the parameter in the parameters of the machine learning model. A single RE may be designated to indicate the sign of the gradient value relative to the parameter in the parameters of the machine learning model as configured by the indication received from the network node 804 at 808. Here, the presence of the analog signal in the single RE may indicate that the sign of the gradient value is positive, and the absence of the analog signal in the RE may indicate that the sign of the gradient value is negative. For example, for positive gradient values ​​(e.g., 1.3 or 0.5), the first UE 802 may send an analog signal representing a value of +1 in a positive RE to indicate that the sign of the gradient value is positive, and for negative gradient values ​​(e.g., -0.4 or -1.2), the UE may defer sending an analog signal in a positive RE to indicate that the sign of the gradient value is negative.

[0137] The network node 804 may obtain an aggregated analog signal in at least one RE designated for a plurality of UEs to indicate a gradient value of the parameter of the machine learning model relative to the plurality of UEs. The obtained aggregated analog signal may represent a plurality of analog signals cumulatively obtained from the plurality of UEs at the network node 804. Here, each of the plurality of analog signals may indicate a gradient value of the parameter in the parameter of the machine learning model calculated at a corresponding UE among the plurality of UEs.

[0138] In a first aspect, the network node 804 may obtain at least one aggregated analog signal in a pair of REs designated for multiple UEs to indicate a gradient value of one parameter in a machine learning model relative to the multiple UEs. The at least one aggregated analog signal obtained in a pair of REs may represent a plurality of analog signals cumulatively obtained from the multiple UEs at the network node 804, each of the plurality of analog signals indicating a gradient value of the parameter in the machine learning model calculated at a corresponding UE in the multiple UEs. The gradient value may include a positive gradient value or a negative gradient value. The at least one aggregated analog signal may include a first aggregated analog signal and a second aggregated analog signal, and the pair of REs designated for indicating the parameter in the parameters of the machine learning model may include a first RE and a second RE. The first aggregated analog signal obtained in the first RE may represent a first group of analog signals cumulatively obtained at the network node 804, each of the first group of analog signals indicating a positive gradient value, and the second aggregated analog signal obtained in the second RE may represent a second group of analog signals cumulatively obtained at the network node 804, each of the second group of analog signals indicating a negative gradient value. For example, the first aggregated analog signal obtained in the first RE may be represented as Where N + It can represent the number of UEs with positive gradient values, can represent the magnitude of the positive gradient value, and may represent an arbitrary phase experienced by a UE with index n due to a time-phase-frequency error; and a second aggregate analog signal obtained in the second RE may be represented as Where N - It can represent the number of UEs with negative gradient values, can represent the magnitude of the positive gradient value, and It can be represented by the time-phase-frequency error caused by the index n - Any phase experienced by the UE.

[0139] In a second aspect, the network node 804 may obtain an aggregate analog signal in a single RE designated for multiple UEs to indicate a sign of a gradient value of one of the parameters of the machine learning model relative to the multiple UEs, the aggregate analog signal obtained in the single RE representing multiple analog signals cumulatively obtained from the multiple UEs at the network node 804. Each of the multiple analog signals may indicate whether a sign of a gradient value of the parameter in the parameters of the machine learning model calculated at a corresponding UE among the multiple UEs is positive or negative. Here, at least one aggregate analog signal is obtained in the single RE based on the indication of the single RE. For example, the aggregate analog signal obtained in the single RE may be represented as Where N +It can represent the number of UEs with positive gradient values, can represent the magnitude of the positive gradient value, and may represent an arbitrary phase experienced by a UE with index n due to a time-phase-frequency error.

[0140] At 820, the network node 804 may compare a first magnitude of the first aggregate analog signal obtained in the first RE with a second magnitude of the second aggregate analog signal obtained in the second RE. + | 2 with |y - | 2 Compare and based on |y + | 2 with |y - | 2 The comparison results are used to determine the aggregation gradient value.

[0141] At 822, the network node 804 may compare the magnitude of the aggregate analog signal obtained in a single RE with a threshold value. That is, the network node 804 may compare |y + | 2 is compared with the threshold and based on |y + | 2 The aggregate gradient value is determined by comparing the result with the threshold. Here, the threshold may be determined or configured by the network node 804.

[0142] At 824, the network node 804 may generate an aggregate gradient of the parameter in the parameter of the machine learning model based on the aggregated simulated signal obtained from the plurality of UEs. Specifically, the aggregate gradient of the parameter of the machine learning may be generated based on the comparison result at 820 or 822.

[0143] In a first aspect, in conjunction with 820, the network node 804 may generate an aggregate gradient of the parameter in the parameters of the machine learning model based on at least one aggregated analog signal obtained in a pair of REs. The aggregate gradient of the parameter in the parameters of the machine learning model may be generated based on the comparison result of the first magnitude and the second magnitude at 820. Based on the first magnitude being greater than the second magnitude, the aggregate gradient may have a value of +p, and based on the first magnitude being less than or equal to the second magnitude, the aggregate gradient has a value of -p, where p is a real number. In one example, based on determining |y + | 2 >|y - | 2 , the network node 804 may declare that the aggregate gradient value is +1. In another example, based on determining |y + | 2 ≤|y - | 2, the server can declare that the aggregate gradient value is -1.

[0144] In a second aspect, in conjunction with 822, the network node 804 may generate an aggregate gradient of the parameter in the parameters of the machine learning model based on the aggregate analog signal obtained in the single RE. The aggregate gradient of the parameter in the parameters of the machine learning model may be generated based on a comparison result of the magnitude of the aggregate analog signal determined at 822 with the threshold. Based on the magnitude being greater than the threshold, the aggregate gradient may be generated to have a value of +p, and based on the magnitude being less than or equal to the threshold, the aggregate gradient may be generated to have a value of -p, where p is a real number. For example, p may be +1. In one example, based on determining |y + | 2 >threshold, the network node 804 may declare the aggregate gradient value to be +1. In another example, based on determining |y + | 2 ≤threshold, the network node 804 may declare the aggregate gradient value to be −1. Here, the threshold may be determined or configured by the network node 804.

[0145] At 826, the network node 804 may update the parameter of the machine learning model based on the aggregated gradient generated at 824. The network node 804 may use the aggregated gradient value generated at 824 to update the parameter of the machine learning model.

[0146] At 828, the network node 804 may send information associated with updated ones of the parameters of the machine learning model to a plurality of UEs including the first UE 802 and the second UE 803, the information associated with the updated parameters being based at least in part on the aggregated simulated signal received at the network node 804. The first UE 802 may receive information associated with updated ones of the parameters of the machine learning model from the network node 804, the information associated with the updated parameters being based at least in part on the simulated signal sent to the network node 804. The updated parameter information may include the updated parameters at 826.

[0147] At 830, the first UE 802 may update the parameter of the machine learning model based on the information associated with the updated parameter received from the network node 804 at 828. The information associated with the updated parameter may include updated parameters (e.g., updated global parameters) for multiple UEs including the first UE 802, and the first UE 802 may replace the original parameter with the updated parameter from the information associated with the updated parameter received from the network node 804.

[0148] Fig. 9900 is a flow chart of a method of wireless communication. The method may be performed by a UE (e.g., UE 104; device 1704). The UE may be configured with a pair of REs designated for the UE to send an analog signal in one of the pair of REs to indicate a gradient value relative to a parameter to a network node. The UE may send an analog signal in one of the pair of REs representing the gradient value of the parameter, receive information associated with updated parameter information from the network node, and update the parameter of the machine learning model based on the updated parameter information received from the network node.

[0149] At 906, the UE may receive parameters of the machine learning model from the network node. Here, the parameters of the machine learning model may be global parameters of the machine learning model, which may be shared by multiple UEs including the UE. For example, at 806, the first UE 802 may receive parameters of the machine learning model from the network node 804. In addition, 906 may be performed by the federated ML parameter reporting component 198.

[0150] At 908, the UE may receive an indication of a pair of REs from the network node 804, the pair of REs being configured for a plurality of UEs including the UE to indicate a gradient value of the parameter relative to the parameter of the machine learning model. The indication may configure a pair of REs for a plurality of UEs including the UE. The pair of REs designated for indicating the parameter of the machine learning model may include: a first RE, the first RE being designated for indicating a positive gradient value; and a second RE, the second RE being designated for indicating a negative gradient value. For example, at 808, the first UE 802 may receive an indication of a pair of REs from the network node 804. In addition, 908 may be performed by the federated ML parameter reporting component 198.

[0151] At 910, the UE may calculate a gradient value relative to one of the parameters of the machine learning model. Here, the gradient value may include a positive gradient value or a negative gradient value. The UE may perform training of the parameter received at 906 independently of other UEs, and calculate the gradient value of the parameter based on the training of the parameter using a training data set. For example, at 810, the first UE 802 may calculate a gradient value relative to one of the parameters of the machine learning model. In addition, 910 may be performed by the federated ML parameter reporting component 198.

[0152] At 914, the UE may send an analog signal indicating the magnitude of the gradient value of the parameter in the parameters of the machine learning model to the network node in one of the pair of REs. A pair of REs may be designated to indicate the gradient value of the parameter in the parameters of the machine learning model as configured by the indication received from the network node at 908. The pair of REs designated to indicate the parameters of the machine learning model may include: a first RE, which is designated to indicate a positive gradient value; and a second RE, which is designated to indicate a negative gradient value. The UE may send the analog signal sent in the first RE or the second RE based on the gradient value. For example, the UE may observe a positive value of +1.3 or a positive value of +0.5 as a given gradient of the parameter, and the UE may send an analog signal representing the value of +1.3 or the value of +0.5 on the first RE. The UE may observe a negative value of -0.4 or a negative value of -1.2 as a given gradient of the parameter, and the UE may send an analog signal representing the value of +0.4 or the value of +1.2 on the second RE. For example, at 814, the first UE 802 can send an analog signal indicating a magnitude of a gradient value relative to one of the parameters of the machine learning model in one of a pair of REs to the network node 804. In addition, 914 can be performed by the federated ML parameter reporting component 198.

[0153] At 928, the UE may receive information associated with updated parameters in the parameters of the machine learning model from the network node, the information associated with the updated parameters being based at least in part on the simulated signal sent to the network node. The updated parameter information may indicate an updated value of the parameter or a gradient value of the parameter. For example, at 826, the first UE 802 may receive information associated with updated parameters in the parameters of the machine learning model from the network node 804, the information associated with the updated parameters being based at least in part on the simulated signal sent to the network node 804. In addition, 926 may be performed by the federated ML parameter reporting component 198.

[0154] At 930, the UE may update the parameter of the machine learning model based on the information associated with the updated parameter received from the network node at 928. The information associated with the updated parameter information may include updated parameters (e.g., updated global parameters) for multiple UEs including the UE, and the UE may replace the original parameter with the updated parameter from the information associated with the updated parameter received from the network node. For example, at 830, the first UE 802 may update the parameter of the machine learning model based on the information associated with the updated parameter received from the network node 804. In addition, 928 may be performed by the federated ML parameter reporting component 198.

[0155] Fig.101000 is a flow chart of a method of wireless communication. The method may be performed by a UE (e.g., UE 104; device 1704). The UE may be configured with a pair of REs designated for the UE to send an analog signal in one of the pair of REs to indicate a gradient value relative to a parameter to a network node. The UE may send an analog signal in one of the pair of REs representing the gradient value of the parameter, receive information associated with updated parameter information from the network node, and update the parameter of the machine learning model based on the updated parameter information received from the network node.

[0156] At 1006, the UE may receive parameters of the machine learning model from the network node. Here, the parameters of the machine learning model may be global parameters of the machine learning model, which may be shared by multiple UEs including the UE. For example, at 806, the first UE 802 may receive parameters of the machine learning model from the network node 804. In addition, 1006 may be performed by the federated ML parameter reporting component 198.

[0157] At 1010, the UE may calculate a gradient value relative to one of the parameters of the machine learning model. Here, the gradient value may include a positive gradient value or a negative gradient value. The UE may perform training of the parameter received at 1006 independently of other UEs, and calculate the gradient value of the parameter based on the training of the parameter using a training data set. For example, at 810, the first UE 802 may calculate a gradient value relative to one of the parameters of the machine learning model. In addition, 1010 may be performed by the federated ML parameter reporting component 198.

[0158] At 1014, the UE may send an analog signal indicating a magnitude of a gradient value of a parameter in a machine learning model relative to the parameter in the parameter of the machine learning model to the network node in one of the pair of REs. A pair of REs may be designated to indicate a gradient value relative to the parameter in the parameter of the machine learning model as configured by the indication received from the network node at 1008. The pair of REs designated to indicate the parameter of the machine learning model may include: a first RE, the first RE designated to indicate a positive gradient value; and a second RE, the second RE designated to indicate a negative gradient value. The UE may send the analog signal sent in the first RE or the second RE based on the gradient value. For example, the UE may observe a positive value of +1.3 or a positive value of +0.5 as a given gradient of the parameter, and the UE may send an analog signal representing a value of +1.3 or a value of +0.5 on the first RE. The UE may observe a negative value of -0.4 or a negative value of -1.2 as a given gradient of the parameter, and the UE may send an analog signal representing a value of +0.4 or a value of +1.2 on the second RE. For example, at 814, the first UE 802 can send an analog signal in one of a pair of REs indicating a magnitude of a gradient value relative to the parameter of the machine learning model to the network node 804. In addition, 1014 can be performed by the federated ML parameter reporting component 198.

[0159] Fig.11 1100 is a flow chart of a method of wireless communication. The method may be performed by a UE (e.g., UE 104; device 1704). The UE may be configured with a single RE designated for the UE to send an analog signal in the single RE to indicate a gradient value relative to a parameter to a network node. The UE may send an analog signal in the single RE representing the gradient value of the parameter, receive information associated with updated parameter information from the network node, and update the parameter of the machine learning model based on the updated parameter information received from the network node.

[0160] At 1106, the UE may receive parameters of the machine learning model from the network node. Here, the parameters of the machine learning model may be global parameters of the machine learning model, which may be shared by multiple UEs including the UE. For example, at 806, the first UE 802 may receive parameters of the machine learning model from the network node 804. In addition, 1106 may be performed by the federated ML parameter reporting component 198.

[0161] At 1108, the UE may receive an indication of a single RE from the network node 804, the single RE being configured for multiple UEs including the UE to indicate a gradient value of the parameter relative to the parameter of the machine learning model. The indication may configure a single RE for multiple UEs including the UE. Here, the single RE may be designated to indicate a positive gradient value or a negative gradient value of the gradient value of the parameter relative to the parameter of the machine learning model. For example, at 808, the first UE 802 may receive an indication of a single RE from the network node 804. In addition, 1108 may be performed by the federated ML parameter reporting component 198.

[0162] At 1110, the UE may calculate a gradient value relative to one of the parameters of the machine learning model. Here, the gradient value may include a positive gradient value or a negative gradient value. The UE may perform training of the parameter received at 1106 independently of other UEs, and calculate the gradient value of the parameter based on the training of the parameter using a training data set. For example, at 810, the first UE 802 may calculate a gradient value relative to one of the parameters of the machine learning model. In addition, 1110 may be performed by the federated ML parameter reporting component 198.

[0163] At 1114, the UE may send an analog signal to the network node in a single RE or skip sending the analog signal based on the sign of the gradient value relative to the parameter in the parameters of the machine learning model. A single RE may be designated to indicate the sign of the gradient value relative to the parameter in the parameters of the machine learning model as configured by the indication received from the network node at 1108. Here, the presence of the analog signal in the single RE may indicate that the sign of the gradient value is positive, and the absence of the analog signal in the RE may indicate that the sign of the gradient value is negative. For example, for a positive gradient value (e.g., 1.3 or 0.5), the UE may send an analog signal representing a value of +1 in a positive RE to indicate that the sign of the gradient value is positive, and for a negative gradient value (e.g., -0.4 or -1.2), the UE may defer sending an analog signal in a positive RE to indicate that the sign of the gradient value is negative. For example, at 814, the first UE 802 may send an analog signal to the network node 804 in a single RE or skip sending the analog signal based on the sign of the gradient value relative to the parameter in the parameters of the machine learning model. Additionally, 1114 can be performed by the federated ML parameter reporting component 198 .

[0164] At 1128, the UE may receive information associated with updated parameters in the parameters of the machine learning model from the network node, the information associated with the updated parameters being based at least in part on the simulated signal sent to the network node. The updated parameter information may indicate an updated value of the parameter or a gradient value of the parameter. For example, at 826, the first UE 802 may receive information associated with updated parameters in the parameters of the machine learning model from the network node 804, the information associated with the updated parameters being based at least in part on the simulated signal sent to the network node 804. In addition, 1126 may be performed by the federated ML parameter reporting component 198.

[0165] At 1130, the UE may update the parameter of the machine learning model based on the information associated with the updated parameter received from the network node at 1128. The information associated with the updated parameter information may include updated parameters (e.g., updated global parameters) for multiple UEs including the UE, and the UE may replace the original parameter with the updated parameter from the information associated with the updated parameter received from the network node. For example, at 830, the first UE 802 may update the parameter of the machine learning model based on the information associated with the updated parameter received from the network node 804. In addition, 1128 may be performed by the federated ML parameter reporting component 198.

[0166] Fig.12 1200 is a flow chart of a method of wireless communication. The method may be performed by a UE (e.g., UE 104; device 1704). The UE may be configured with a single RE designated for the UE to send an analog signal in the single RE to indicate a gradient value relative to a parameter to a network node. The UE may send an analog signal in the single RE representing the gradient value of the parameter, receive information associated with updated parameter information from the network node, and update the parameter of the machine learning model based on the updated parameter information received from the network node.

[0167] At 1206, the UE may receive parameters of the machine learning model from the network node. Here, the parameters of the machine learning model may be global parameters of the machine learning model, which may be shared by multiple UEs including the UE. For example, at 806, the first UE 802 may receive parameters of the machine learning model from the network node 804. In addition, 1206 may be performed by the federated ML parameter reporting component 198.

[0168] At 1210, the UE may calculate a gradient value relative to one of the parameters of the machine learning model. Here, the gradient value may include a positive gradient value or a negative gradient value. The UE may perform training of the parameter received at 1206 independently of other UEs, and calculate the gradient value of the parameter based on the training of the parameter using a training data set. For example, at 810, the first UE 802 may calculate a gradient value relative to one of the parameters of the machine learning model. In addition, 1210 may be performed by the federated ML parameter reporting component 198.

[0169] At 1214, the UE may send an analog signal to the network node in a single RE or skip sending the analog signal based on the sign of the gradient value relative to the parameter in the parameters of the machine learning model. A single RE may be designated to indicate the sign of the gradient value relative to the parameter in the parameters of the machine learning model as configured by the indication received from the network node at 1208. Here, the presence of the analog signal in the single RE may indicate that the sign of the gradient value is positive, and the absence of the analog signal in the RE may indicate that the sign of the gradient value is negative. For example, for a positive gradient value (e.g., 1.3 or 0.5), the UE may send an analog signal representing a value of +1 in a positive RE to indicate that the sign of the gradient value is positive, and for a negative gradient value (e.g., -0.4 or -1.2), the UE may defer sending an analog signal in a positive RE to indicate that the sign of the gradient value is negative. For example, at 814, the first UE 802 may send an analog signal to the network node 804 in a single RE or skip sending the analog signal based on the sign of the gradient value relative to the parameter in the parameters of the machine learning model. Additionally, 1214 may be performed by the federated ML parameter reporting component 198 .

[0170] Fig.13 1300 is a flowchart of a method of wireless communication. The method may be performed by a base station (e.g., base station 102; network entity 1802 / 1960). A network node may configure a pair of REs for a plurality of UEs including a UE, the pair of REs being designated for the plurality of UEs to send a plurality of analog signals to indicate a gradient value relative to a parameter to the network node. The network node may obtain at least one aggregated analog signal representing the plurality of analog signals, generate an aggregated gradient of the parameter based on the obtained at least one aggregated analog signal, and update the parameter based on the aggregated gradient. The network node may send information associated with an updated parameter of a machine learning model to a plurality of UEs including the UE, and update the parameter of the machine learning model based on the updated parameter information.

[0171] At 1306, the network node may send parameters of the machine learning model for the plurality of UEs. Here, the parameters of the machine learning model may be global parameters of the machine learning model, which may be shared by the plurality of UEs including the UE. For example, at 806, the network node 804 may send parameters of the machine learning model for the plurality of UEs including the first UE 802 and the second UE 803. In addition, 1306 may be performed by the federated ML parameter processing component 199.

[0172] At 1308, the network node may send an indication of a pair of REs to multiple UEs including the UE to configure the multiple UEs to indicate a gradient value of the parameter relative to the parameter of the machine learning model. The indication may configure a pair of REs for multiple UEs. The pair of REs designated for indicating the parameters of the machine learning model may include: a first RE, which is designated to indicate a positive gradient value; and a second RE, which is designated to indicate a negative gradient value. For example, at 808, the network node 804 may send an indication of a pair of REs to multiple UEs including a first UE 802 and a second UE 803 to configure the first UE 802 and the second UE 803 to indicate a gradient value of the parameter relative to the parameter of the machine learning model. In addition, 1308 may be performed by the federal ML parameter processing component 199.

[0173] At 1314, the network node may obtain at least one aggregated analog signal in a pair of REs designated for multiple UEs to indicate a gradient value of one parameter in the parameters of the machine learning model of the multiple UEs. The obtained aggregated analog signal may represent multiple analog signals cumulatively obtained from multiple UEs at the network node. Here, each of the multiple analog signals may indicate a gradient value of the parameter in the parameters of the machine learning model calculated at a corresponding UE in the multiple UEs. At least one aggregated analog signal obtained in a pair of REs may represent multiple analog signals cumulatively obtained from multiple UEs at the network node, each of the multiple analog signals indicating a gradient value of the parameter in the parameters of the machine learning model calculated at a corresponding UE in the multiple UEs. The gradient value may include a positive gradient value or a negative gradient value. The at least one aggregated analog signal may include a first aggregated analog signal and a second aggregated analog signal, and the pair of REs designated for indicating the parameter in the parameters of the machine learning model may include a first RE and a second RE. The first aggregate analog signal obtained in the first RE may represent a first set of analog signals cumulatively obtained at the network node, each analog signal in the first set of analog signals indicating a positive gradient value, and the second aggregate analog signal obtained in the second RE may represent a second set of analog signals cumulatively obtained at the network node, each analog signal in the second set of analog signals indicating a negative gradient value. For example, the first aggregate analog signal obtained in the first RE may be represented as Where N + It can represent the number of UEs with positive gradient values, can represent the magnitude of the positive gradient value, and may represent an arbitrary phase experienced by a UE with index n due to a time-phase-frequency error; and a second aggregate analog signal obtained in the second RE may be represented as Where N _ It can represent the number of UEs with negative gradient values, can represent the magnitude of the positive gradient value, and It can be represented by the time-phase-frequency error due to the index n _ For example, at 814, the network node 804 may obtain at least one aggregated analog signal in a pair of REs designated for multiple UEs to indicate a gradient value of one parameter of a machine learning model relative to the multiple UEs. In addition, 1314 may be performed by the federated ML parameter processing component 199.

[0174] At 1320, the network node may compare a first magnitude of the first aggregate analog signal obtained in the first RE with a second magnitude of the second aggregate analog signal obtained in the second RE. + | 2 with |y _ | 2 Compare and based on |y + | 2 with |y _ | 2 For example, at 820, the network node 804 may compare a first magnitude of a first aggregate analog signal obtained in the first RE with a second magnitude of a second aggregate analog signal obtained in the second RE. In addition, 1320 may be performed by the federated ML parameter processing component 199.

[0175] At 1324, the network node may generate an aggregate gradient of the parameter in the parameters of the machine learning model based on at least one aggregated analog signal obtained in a pair of REs. The aggregate gradient of the parameter in the parameters of the machine learning model may be generated based on the comparison result of the first magnitude and the second magnitude at 1320. Based on the first magnitude being greater than the second magnitude, the aggregate gradient may have a value of +p, and based on the first magnitude being less than or equal to the second magnitude, the aggregate gradient may have a value of -p, where p is a real number. For example, p may be +1. In one example, based on determining |y + | 2 >|y _ | 2, the network node may declare that the aggregate gradient value is +1. In another example, based on determining |y + | 2 ≤|y _ | 2 , the server may declare that the aggregate gradient value is -1. For example, at 824, the network node 804 may generate an aggregate gradient of the parameter in the parameter of the machine learning model based on at least one aggregated simulation signal obtained in a pair of REs. In addition, 1324 may be performed by the federated ML parameter processing component 199.

[0176] At 1326, the network node may update the parameter of the machine learning model based on the aggregated gradient generated at 1324. The network node may update the parameter of the machine learning model using the aggregated gradient value generated at 1324. For example, at 830, the network node 804 may update the parameter of the machine learning model based on the aggregated gradient generated at 824. In addition, 1330 may be performed by the federated ML parameter processing component 199.

[0177] At 1328, the network node may send information associated with updated parameters in the parameters of the machine learning model to multiple UEs, the information associated with the updated parameters being based at least in part on the aggregated simulated signal received at the network node. The first UE may receive information associated with updated parameters in the parameters of the machine learning model from the network node, the information associated with the updated parameters being based at least in part on the simulated signal sent to the network node. The updated parameter information may include the updated parameters at 1326. For example, at 826, the network node 804 may send information associated with updated parameters in the parameters of the machine learning model to multiple UEs including the first UE 802 and the second UE 803, the information associated with the updated parameters being based at least in part on the aggregated simulated signal received at the network node 804. In addition, 1326 may be performed by the federated ML parameter processing component 199.

[0178] Fig.14 1400 is a flowchart of a method of wireless communication. The method may be performed by a base station (e.g., base station 102; network entity 1802 / 1960). A network node may configure a pair of REs for a plurality of UEs including a UE, the pair of REs being designated for the plurality of UEs to send a plurality of analog signals to indicate a gradient value relative to a parameter to the network node. The network node may obtain at least one aggregated analog signal representing the plurality of analog signals, generate an aggregated gradient of the parameter based on the obtained at least one aggregated analog signal, and update the parameter based on the aggregated gradient. The network node may send information associated with an updated parameter of a machine learning model to a plurality of UEs including the UE, and update the parameter of the machine learning model based on the updated parameter information.

[0179] At 1406, the network node may send parameters of the machine learning model for the plurality of UEs. Here, the parameters of the machine learning model may be global parameters of the machine learning model, which may be shared by the plurality of UEs including the UE. For example, at 806, the network node 804 may send parameters of the machine learning model for the plurality of UEs including the first UE 802 and the second UE 803. In addition, 1406 may be performed by the federated ML parameter processing component 199.

[0180] At 1414, the network node may obtain at least one aggregated analog signal in a pair of REs designated for multiple UEs to indicate a gradient value of one parameter in the parameters of the machine learning model of the multiple UEs. The obtained aggregated analog signal may represent multiple analog signals cumulatively obtained from multiple UEs at the network node. Here, each of the multiple analog signals may indicate a gradient value of the parameter in the parameters of the machine learning model calculated at a corresponding UE in the multiple UEs. At least one aggregated analog signal obtained in a pair of REs may represent multiple analog signals cumulatively obtained from multiple UEs at the network node, each of the multiple analog signals indicating a gradient value of the parameter in the parameters of the machine learning model calculated at a corresponding UE in the multiple UEs. The gradient value may include a positive gradient value or a negative gradient value. The at least one aggregated analog signal may include a first aggregated analog signal and a second aggregated analog signal, and the pair of REs designated for indicating the parameter in the parameters of the machine learning model may include a first RE and a second RE. The first aggregate analog signal obtained in the first RE may represent a first set of analog signals cumulatively obtained at the network node, each analog signal in the first set of analog signals indicating a positive gradient value, and the second aggregate analog signal obtained in the second RE may represent a second set of analog signals cumulatively obtained at the network node, each analog signal in the second set of analog signals indicating a negative gradient value. For example, the first aggregate analog signal obtained in the first RE may be represented as Where N + It can represent the number of UEs with positive gradient values, can represent the magnitude of the positive gradient value, and may represent an arbitrary phase experienced by a UE with index n due to a time-phase-frequency error; and a second aggregate analog signal obtained in the second RE may be represented as Where N _ It can represent the number of UEs with negative gradient values, can represent the magnitude of the positive gradient value, and It can be represented by the time-phase-frequency error caused by the index n _For example, at 814, the network node 804 may obtain at least one aggregated analog signal in a pair of REs designated for multiple UEs to indicate a gradient value of one parameter of a machine learning model relative to the multiple UEs. In addition, 1414 may be performed by the federated ML parameter processing component 199.

[0181] At 1426, the network node may update the parameter of the machine learning model based on the aggregated gradient generated at 1424. The network node may update the parameter of the machine learning model using the aggregated gradient value generated at 1424. For example, at 830, the network node 804 may update the parameter of the machine learning model based on the aggregated gradient generated at 824. In addition, 1430 may be performed by the federated ML parameter processing component 199.

[0182] Fig.15 1500 is a flowchart of a method of wireless communication. The method may be performed by a base station (e.g., base station 102; network entity 1802 / 1960). A network node may configure a single RE for multiple UEs including a UE, and the single RE is designated for multiple UEs to send multiple analog signals to indicate a gradient value relative to a parameter to the network node. The network node may obtain at least one aggregated analog signal representing the multiple analog signals, generate an aggregated gradient of the parameter based on the obtained at least one aggregated analog signal, and update the parameter based on the aggregated gradient. The network node may send information associated with an updated parameter of a machine learning model to multiple UEs including the UE, and update the parameter of the machine learning model based on the updated parameter information.

[0183] At 1506, the network node may send parameters of the machine learning model for the plurality of UEs. Here, the parameters of the machine learning model may be global parameters of the machine learning model, which may be shared by the plurality of UEs including the UE. For example, at 806, the network node 804 may send parameters of the machine learning model for the plurality of UEs including the first UE 802 and the second UE 803. In addition, 1506 may be performed by the federated ML parameter processing component 199.

[0184] At 1508, the network node may send an indication of a single RE to multiple UEs including the UE to configure the multiple UEs to indicate a gradient value of the parameter in the parameters of the machine learning model. The indication may configure a single RE for multiple UEs. Here, a single RE may be designated to indicate a positive gradient value or a negative gradient value of a gradient value of the parameter in the parameters of the machine learning model. For example, at 808, the network node 804 may send an indication of a single RE to multiple UEs including a first UE 802 and a second UE 803 to configure the first UE 802 and the second UE 803 to indicate a gradient value of the parameter in the parameters of the machine learning model. In addition, 1508 may be performed by the federated ML parameter processing component 199.

[0185] At 1514, the network node may obtain an aggregated analog signal in a single RE designated for multiple UEs to indicate a sign of a gradient value of one of the parameters of the machine learning model relative to the multiple UEs. The obtained aggregated analog signal may represent multiple analog signals cumulatively obtained from the multiple UEs at the network node. Here, each of the multiple analog signals may indicate a sign of a gradient value of the parameter in the parameters of the machine learning model calculated at a corresponding UE in the multiple UEs. The aggregated analog signal obtained in the single RE represents multiple analog signals cumulatively obtained from the multiple UEs at the network node. Each of the multiple analog signals may indicate whether the sign of the gradient value of the parameter in the parameters of the machine learning model calculated at a corresponding UE in the multiple UEs is one of positive or negative. Here, at least one aggregated analog signal is obtained in the single RE based on an indication of the single RE. For example, the aggregated analog signal obtained in the single RE may be represented as Where N + It can represent the number of UEs with positive gradient values, can represent the magnitude of the positive gradient value, and May represent an arbitrary phase experienced by a UE with index n due to a time-phase-frequency error. For example, at 814, the network node 804 may obtain an aggregated analog signal in a single RE designated for multiple UEs to indicate a gradient value of one of the parameters of the machine learning model relative to the multiple UEs. In addition, 1514 may be performed by the federated ML parameter processing component 199.

[0186] At 1522, the network node may compare the magnitude of the aggregate analog signal obtained in a single RE with a threshold value. That is, the network node may compare |y + | 2 is compared with the threshold and based on |y + | 2The aggregate gradient value is determined by comparing the result with the threshold. Here, the threshold can be determined or configured by the network node. For example, at 822, the network node 804 can compare the magnitude of the aggregate analog signal obtained in a single RE with the threshold. In addition, 1522 can be performed by the federated ML parameter processing component 199.

[0187] At 1524, the network node may generate an aggregate gradient of the parameter in the parameters of the machine learning model based on the aggregate analog signal obtained in the single RE. The aggregate gradient of the parameter in the parameters of the machine learning model may be generated based on the comparison result of the magnitude of the aggregate analog signal determined at 1522 with the threshold. Based on the magnitude being greater than the threshold, the aggregate gradient may have a value of +p, and based on the magnitude being less than or equal to the threshold, the aggregate gradient may have a value of -p, where p is a real number. For example, p may be +1. In one example, based on determining |y + | 2 >threshold, the network node may declare the aggregate gradient value to be +1. In another example, based on determining |y + | 2 ≤threshold, the network node may declare that the aggregate gradient value is -1. Here, the threshold may be determined or configured by the network node. For example, at 824, the network node 804 may generate an aggregate gradient of the parameter in the parameters of the machine learning model based on the aggregated simulation signal obtained in the single RE. In addition, 1524 may be performed by the federated ML parameter processing component 199.

[0188] At 1526, the network node may update the parameter of the machine learning model based on the aggregated gradient generated at 1524. The network node may update the parameter of the machine learning model using the aggregated gradient value generated at 1324. For example, at 830, the network node 804 may update the parameter of the machine learning model based on the aggregated gradient generated at 824. In addition, 1530 may be performed by the federated ML parameter processing component 199.

[0189] At 1528, the network node may send information associated with updated parameters in the parameters of the machine learning model to multiple UEs, the information associated with the updated parameters being based at least in part on the aggregated simulated signal received at the network node. The first UE may receive information associated with updated parameters in the parameters of the machine learning model from the network node, the information associated with the updated parameters being based at least in part on the simulated signal sent to the network node. The updated parameter information may include the updated parameters at 1526. For example, at 826, the network node 804 may send information associated with updated parameters in the parameters of the machine learning model to multiple UEs including the first UE 802 and the second UE 803, the information associated with the updated parameters being based at least in part on the aggregated simulated signal received at the network node 804. In addition, 1526 may be performed by the federated ML parameter processing component 199.

[0190] Fig.16 1600 is a flowchart of a method of wireless communication. The method may be performed by a base station (e.g., base station 102; network entity 1802 / 1960). A network node may configure a single RE for multiple UEs including a UE, and the single RE is designated for multiple UEs to send multiple analog signals to indicate a gradient value relative to a parameter to the network node. The network node may obtain at least one aggregated analog signal representing the multiple analog signals, generate an aggregated gradient of the parameter based on the obtained at least one aggregated analog signal, and update the parameter based on the aggregated gradient. The network node may send information associated with an updated parameter of a machine learning model to multiple UEs including the UE, and update the parameter of the machine learning model based on the updated parameter information.

[0191] At 1606, the network node may send parameters of the machine learning model for the plurality of UEs. Here, the parameters of the machine learning model may be global parameters of the machine learning model, which may be shared by the plurality of UEs including the UE. For example, at 806, the network node 804 may send parameters of the machine learning model for the plurality of UEs including the first UE 802 and the second UE 803. In addition, 1606 may be performed by the federated ML parameter processing component 199.

[0192] At 1614, the network node may obtain an aggregate analog signal in a single RE designated for multiple UEs to indicate a sign of a gradient value of one of the parameters of the machine learning model relative to the multiple UEs. The obtained aggregate analog signal may represent multiple analog signals cumulatively obtained from the multiple UEs at the network node. Here, each of the multiple analog signals may indicate a sign of a gradient value of the parameter in the parameters of the machine learning model calculated at a corresponding UE in the multiple UEs. The aggregate analog signal obtained in the single RE represents multiple analog signals cumulatively obtained from the multiple UEs at the network node. Each of the multiple analog signals may indicate whether the sign of the gradient value of the parameter in the parameters of the machine learning model calculated at a corresponding UE in the multiple UEs is one of positive or negative. Here, at least one aggregate analog signal is obtained in the single RE based on an indication of the single RE. For example, the aggregate analog signal obtained in the single RE may be represented as Where N + It can represent the number of UEs with positive gradient values, can represent the magnitude of the positive gradient value, and May represent an arbitrary phase experienced by a UE with index n due to a time-phase-frequency error. For example, at 814, the network node 804 may obtain an aggregated analog signal in a single RE designated for multiple UEs to indicate a gradient value of one of the parameters of the machine learning model relative to the multiple UEs. In addition, 1614 may be performed by the federated ML parameter processing component 199.

[0193] At 1626, the network node may update the parameter of the machine learning model based on the aggregated gradient generated at 1624. The network node may update the parameter of the machine learning model using the aggregated gradient value generated at 1324. For example, at 830, the network node 804 may update the parameter of the machine learning model based on the aggregated gradient generated at 824. In addition, 1630 may be performed by the federated ML parameter processing component 199.

[0194] Fig.171700 is a diagram illustrating an example of a hardware implementation for an apparatus 1704. The apparatus 1704 may be a UE, a component of a UE, or may implement UE functionality. In some aspects, the apparatus 1304 may include a cellular baseband processor 1724 (also referred to as a modem) coupled to one or more transceivers 1722 (e.g., a cellular RF transceiver). The cellular baseband processor 1724 may include on-chip memory 1724'. In some aspects, the apparatus 1704 may also include one or more subscriber identity module (SIM) cards 1720 and an application processor 1706 coupled to a secure digital (SD) card 1708 and a screen 1710. The application processor 1706 may include on-chip memory 1706'. In some aspects, the device 1704 may also include a Bluetooth module 1712, a WLAN module 1714, an SPS module 1716 (e.g., a GNSS module), one or more sensor modules 1718 (e.g., an atmospheric pressure sensor / altimeter; a motion sensor such as an inertial management unit (IMU), a gyroscope, and / or an accelerometer; light detection and ranging (LIDAR), radio-aided detection and ranging (RADAR), sound navigation and ranging (SONAR), a magnetometer, audio, and / or other technologies for positioning), an additional memory module 1726, a power source 1730, and / or a camera 1732. The Bluetooth module 1712, the WLAN module 1714, and the SPS module 1716 may include an on-chip transceiver (TRX) (or in some cases, only a receiver (RX)). The Bluetooth module 1712, the WLAN module 1714, and the SPS module 1716 may include their own dedicated antennas and / or communicate using antennas 1780. The cellular baseband processor 1724 communicates with the UE 104 and / or with the RU associated with the network entity 1702 through the transceiver 1722 via one or more antennas 1780. The cellular baseband processor 1724 and the application processor 1706 may each include a computer-readable medium / memory 1724', 1706', respectively. The additional memory module 1726 may also be considered as a computer-readable medium / memory. Each computer-readable medium / memory 1724', 1706', 1726 may be non-transitory. The cellular baseband processor 1724 and the application processor 1706 are each responsible for general processing, including executing software stored on a computer-readable medium / memory. The software enables the cellular baseband processor 1724 / application processor 1706 to perform the various functions described above when executed by the cellular baseband processor 1724 / application processor 1706. The computer-readable medium / memory can also be used to store data manipulated by the cellular baseband processor 1724 / application processor 1706 when executing the software.The cellular baseband processor 1724 / application processor 1706 may be a component of the UE 350 and may include the memory 360 and / or at least one of the TX processor 368, the RX processor 356, and the controller / processor 359. In one configuration, the device 1704 may be a processor chip (modem and / or application) and include only the cellular baseband processor 1724 and / or the application processor 1706, while in another configuration, the device 1704 may be the entire UE (e.g., see. Figure 3 350) and includes additional modules of device 1704.

[0195] As discussed above, the federated ML parameter reporting component 198 is configured to: receive parameters of a machine learning model from a network node; calculate a gradient value relative to one of the parameters of the machine learning model, the gradient value comprising a positive gradient value or a negative gradient value; and send an analog signal indicating the magnitude of the gradient value relative to the parameter in the parameters of the machine learning model to the network node in one of a pair of REs, the pair of REs being designated for indicating the gradient value relative to the parameter in the parameters of the machine learning model. The federated ML parameter reporting component 198 is also configured to: receive parameters of a machine learning model from a network node; calculate a gradient value relative to one of the parameters of the machine learning model, the sign of the gradient value being one of positive or negative; and send the analog signal to the network node in a single RE, or skip sending the analog signal, based on the sign of the gradient value relative to the parameter in the parameters of the machine learning model, the single RE being designated for indicating the sign of the gradient value relative to the parameter in the parameters of the machine learning model. The federated ML parameter reporting component 198 may be within the cellular baseband processor 1724, within the application processor 1706, or within both the cellular baseband processor 1724 and the application processor 1706. The federated ML parameter reporting component 198 may be one or more hardware components specifically configured to implement the process / algorithm, implemented by one or more processors configured to execute the process / algorithm, stored in a computer-readable medium for implementation by one or more processors, or some combination thereof. As shown, the device 1704 may include a variety of components configured for various functions. In one configuration, the device 1704 (and specifically, the cellular baseband processor 1724 and / or the application processor 1706) includes: a component for receiving a parameter of a machine learning model from a network node; a component for calculating a gradient value relative to one of the parameters of the machine learning model, the gradient value including a positive gradient value or a negative gradient value; and a component for sending an analog signal indicating the magnitude of the gradient value of the parameter relative to the parameter of the machine learning model to the network node in one of a pair of REs, the pair of REs being designated for indicating the gradient value relative to the parameter of the machine learning model. In one configuration, a pair of REs designated for indicating parameters of a machine learning model includes a first RE designated for indicating a positive gradient value and a second RE designated for indicating a negative gradient value, wherein an analog signal is sent in the first RE or the second RE based on the gradient value.In one configuration, the device 1704 (and specifically, the cellular baseband processor 1724 and / or the application processor 1706) also includes: a component for receiving information associated with updated parameters of a machine learning model from a network node, the updated parameters being based at least in part on an analog signal sent to the network node; and a component for updating the parameters of the machine learning model based on the information associated with the updated parameters received from the network node. In one configuration, the information associated with the updated parameters includes updated parameters based at least in part on an analog signal sent to the network node. In one configuration, the device 1704 (and specifically, the cellular baseband processor 1724 and / or the application processor 1706) also includes a component for receiving an indication of a pair of REs from a network node, wherein the analog signal is sent in one of the pair of REs based on the indication of the pair of REs received from the network node. In one configuration, the apparatus 1704 (and specifically, the cellular baseband processor 1724 and / or the application processor 1706) includes: means for receiving parameters of a machine learning model from a network node; means for calculating a gradient value with respect to one of the parameters of the machine learning model, the sign of the gradient value being one of positive or negative; and means for sending an analog signal to the network node in a single RE or skipping the sending of the analog signal based on the sign of the gradient value with respect to the parameter in the parameters of the machine learning model, the single RE being designated for indicating the sign of the gradient value with respect to the parameter in the parameters of the machine learning model. In one configuration, the presence of the analog signal in the single RE indicates that the sign of the gradient value is positive, and the absence of the analog signal in the RE indicates that the sign of the gradient value is negative. In one configuration, the device 1704 (and specifically, the cellular baseband processor 1724 and / or the application processor 1706) also includes: a component for receiving information associated with an updated parameter of the machine learning model from a network node, the updated parameter being based at least in part on an analog signal sent to the network node; and a component for updating the parameter of the machine learning model based on the information associated with the updated parameter received from the network node. In one configuration, the information associated with the updated parameter includes an updated parameter based at least in part on an analog signal sent to the network node. In one configuration, the device 1704 (and specifically, the cellular baseband processor 1724 and / or the application processor 1706) also includes a component for receiving an indication of a single RE from the network node, wherein the analog signal is sent in the single RE based on the indication of the single RE received from the network node. The component can be a federated ML parameter reporting component 198 of the device 1704 configured to perform the functions recorded by the component. As described above, the device 1704 may include the TX processor 368, the RX processor 356, and the controller / processor 359.Thus, in one configuration, the means may be the TX processor 368, the RX processor 356, and / or the controller / processor 359 configured to perform the functions recited by the means.

[0196] Fig.18 1800 is a diagram illustrating an example of a hardware implementation for a network entity 1802. The network entity 1802 may be a BS, may be a component of a BS, or may implement BS functionality. The network entity 1802 may include at least one of a CU 1810, a DU 1830, or a RU 1840. For example, depending on the layer functionality handled by the federated ML parameter processing component 199, the network entity 1802 may include a CU 1810; both a CU 1810 and a DU 1830; each of the CU 1810, the DU 1830, and the RU 1840; the DU 1830; both a DU 1830 and a RU 1840; or the RU 1840. The CU 1810 may include a CU processor 1812. The CU processor 1812 may include an on-chip memory 1812'. In some aspects, the CU 1810 may also include an additional memory module 1814 and a communication interface 1818. CU 1810 communicates with DU 1830 via a midhaul link, such as an F1 interface. DU 1830 may include a DU processor 1832. DU processor 1832 may include on-chip memory 1832'. In some aspects, DU 1830 may also include additional memory modules 1834 and a communication interface 1838. DU 1830 communicates with RU 1840 via a fronthaul link. RU 1840 may include a RU processor 1842. RU processor 1842 may include on-chip memory 1842'. In some aspects, RU 1840 may also include additional memory modules 1844, one or more transceivers 1846, an antenna 1880, and a communication interface 1848. RU 1840 communicates with UE 104. On-chip memory 1812', 1832', 1842' and additional memory modules 1814, 1834, 1844 may each be considered a computer-readable medium / memory. Each computer readable medium / memory may be non-transitory. Each of processors 1812, 1832, 1842 is responsible for general processing, including executing software stored on the computer readable medium / memory. The software, when executed by the corresponding processor, causes the processor to perform the various functions described above. The computer readable medium / memory may also be used to store data manipulated by the processor when executing the software.

[0197] As discussed above, the federated ML parameter processing component 199 is configured to: output parameters of a machine learning model for multiple UEs for transmission; obtain at least one aggregated analog signal in a pair of REs designated for multiple UEs to indicate a gradient value of a parameter in the parameters of the machine learning model for multiple UEs, the at least one aggregated analog signal obtained in a pair of REs representing multiple analog signals cumulatively obtained from multiple UEs at a network node, each analog signal of the multiple analog signals indicating a gradient value of the parameter in the parameters of the machine learning model calculated at a corresponding UE of the multiple UEs, the gradient value including a positive gradient value or a negative gradient value; and update the parameter in the parameters of the machine learning model based on the at least one aggregated analog signal obtained in a pair of REs. The federated ML parameter processing component 199 is also configured to: output parameters of the machine learning model for multiple UEs for transmission; obtain an aggregated analog signal in a single RE designated for multiple UEs to indicate a sign of a gradient value relative to one of the parameters of the machine learning model for multiple UEs, the aggregated analog signal obtained in the single RE representing multiple analog signals cumulatively obtained from multiple UEs at the network node, each of the multiple analog signals indicating a sign of a gradient value relative to the parameter in the parameters of the machine learning model calculated at the corresponding UE of the multiple UEs as one of positive or negative; and update the parameter in the parameters of the machine learning model based on the aggregated analog signal obtained in the single RE. The federated ML parameter processing component 199 may be within one or more processors of one or more of the CU 1810, DU 1830, and RU 1840. The federated ML parameter processing component 199 may be one or more hardware components specifically configured to implement the process / algorithm, implemented by one or more processors configured to execute the process / algorithm, stored in a computer-readable medium for implementation by one or more processors, or some combination thereof. The network entity 1802 may include a variety of components configured for various functions. In one configuration, the network entity 1802 includes: a component for outputting parameters of a machine learning model for multiple UEs for transmission; a component for obtaining at least one aggregated analog signal in a pair of REs designated for multiple UEs to indicate a gradient value of a parameter in the parameters of the machine learning model for multiple UEs, the at least one aggregated analog signal obtained in a pair of REs representing multiple analog signals cumulatively obtained from multiple UEs at a network node, each of the multiple analog signals indicating a gradient value of the parameter in the parameters of the machine learning model calculated at a corresponding UE of the multiple UEs, the gradient value including a positive gradient value or a negative gradient value; and a component for updating the parameter in the parameters of the machine learning model based on the at least one aggregated analog signal obtained in a pair of REs.In one configuration, at least one aggregated analog signal includes a first aggregated analog signal and a second aggregated analog signal, and a pair of REs designated for indicating the parameter in the parameter of the machine learning model includes the first RE and the second RE; the first aggregated analog signal obtained in the first RE represents a first group of analog signals cumulatively obtained at the network node, each analog signal in the first group of analog signals indicating a positive gradient value; and the second aggregated analog signal obtained in the second RE represents a second group of analog signals cumulatively obtained at the network node, each analog signal in the second group of analog signals indicating a negative gradient value. In one configuration, the network entity 1802 also includes a component for comparing a first magnitude of the first aggregated analog signal obtained in the first RE with a second magnitude of the second aggregated analog signal obtained in the second RE, and wherein the parameter is updated based on the comparison result of the first magnitude with the second magnitude. In one configuration, the network entity 1802 also includes a component for generating an aggregated gradient of the parameter based on the comparison result of the first magnitude with the second magnitude, and wherein the parameter is updated based on the aggregated gradient of the parameter. In one configuration, based on the first magnitude being greater than the second magnitude, the aggregate gradient of the parameter has a value of +p, based on the first magnitude being less than or equal to the second magnitude, the aggregate gradient of the parameter has a value of -p, and p is a real number. In one configuration, the network entity 1802 also includes a component for outputting information associated with updated parameters of the machine learning model to multiple UEs for transmission. In one configuration, the information associated with the updated parameters includes the updated parameters. In one configuration, the network entity 1802 also includes a component for outputting an indication of a pair of REs for multiple UEs for transmission, and wherein at least one aggregated analog signal is obtained in the pair of REs based on the indication of the pair of REs. In one configuration, the network entity 1802 includes: a component for outputting parameters of a machine learning model for multiple UEs for transmission; a component for obtaining an aggregated analog signal in a single RE designated for multiple UEs to indicate a sign of a gradient value relative to one of the parameters of the machine learning model for multiple UEs, the aggregated analog signal obtained in the single RE representing multiple analog signals cumulatively obtained from multiple UEs at a network node, each of the multiple analog signals indicating a sign of a gradient value relative to the parameter in the parameters of the machine learning model calculated at a corresponding UE of the multiple UEs as one of positive or negative; and a component for updating the parameter in the parameters of the machine learning model based on the aggregated analog signal obtained in the single RE. In one configuration, the presence of the analog signal in the single RE indicates that the sign of the gradient value is positive, and the absence of the analog signal in the single RE indicates that the sign of the gradient value is negative. In one configuration, the network entity 1802 also includes a component for comparing the magnitude of the aggregated analog signal obtained in the single RE with a threshold value, wherein the parameter is updated based on the comparison result of the magnitude with the threshold value.In one configuration, the network entity 1802 also includes a component for generating an aggregate gradient of the parameter based on a comparison result of the magnitude with a threshold, wherein the parameter is updated based on the aggregate gradient of the parameter. In one configuration, based on the magnitude being greater than the threshold, the aggregate gradient of the parameter has a value of +p, based on the magnitude being less than or equal to the threshold, the aggregate gradient of the parameter has a value of -p, and p is a real number. In one configuration, the network entity 1802 also includes a component for outputting information associated with updated parameters of the machine learning model to multiple UEs for transmission. In one configuration, the information associated with the updated parameters includes the updated parameters. In one configuration, the network entity 1802 also includes a component for outputting an indication of a single RE for multiple UEs for transmission, wherein at least one aggregated analog signal is obtained in a single RE based on the indication of the single RE. The component may be a federated ML parameter processing component 199 of the network entity 1802 configured to perform the functions recorded by the component. As described above, the network entity 1802 may include a TX processor 316, an RX processor 370, and a controller / processor 375. Thus, in one configuration, the means may be the TX processor 316, the RX processor 370, and / or the controller / processor 375 configured to perform the functions recited by the means.

[0198] Fig.19 1900 is an example of a hardware implementation for a network entity 1960. In one example, the network entity 1960 may be within the core network 120. The network entity 1960 may include a network processor 1912. The network processor 1912 may include an on-chip memory 1912'. In some aspects, the network entity 1960 may also include an additional memory module 1914. The network entity 1960 communicates with the CU 1902 directly (e.g., backhaul link) or indirectly (e.g., through the RIC) via a network interface 1980. The on-chip memory 1912' and the additional memory module 1914 may each be considered as a computer-readable medium / memory. Each computer-readable medium / memory may be non-transitory. The processor 1912 is responsible for general processing, including executing software stored on a computer-readable medium / memory. The software, when executed by the corresponding processor, enables the processor to perform the various functions described above. The computer-readable medium / memory may also be used to store data manipulated by the processor when executing the software.

[0199] As discussed above, the federated ML parameter processing component 199 is configured to: output parameters of a machine learning model for multiple UEs for transmission; obtain at least one aggregated analog signal in a pair of REs designated for multiple UEs to indicate a gradient value of a parameter in the parameters of the machine learning model for multiple UEs, the at least one aggregated analog signal obtained in a pair of REs representing multiple analog signals cumulatively obtained from multiple UEs at a network node, each analog signal of the multiple analog signals indicating a gradient value of the parameter in the parameters of the machine learning model calculated at a corresponding UE of the multiple UEs, the gradient value including a positive gradient value or a negative gradient value; and update the parameter in the parameters of the machine learning model based on the at least one aggregated analog signal obtained in a pair of REs. The federated ML parameter processing component 199 is also configured to: output parameters of the machine learning model for multiple UEs for transmission; obtain an aggregated analog signal in a single RE designated for multiple UEs to indicate a sign of a gradient value relative to one of the parameters of the machine learning model for multiple UEs, the aggregated analog signal obtained in the single RE representing multiple analog signals cumulatively obtained from multiple UEs at the network node, each of the multiple analog signals indicating a sign of a gradient value relative to the parameter in the parameters of the machine learning model calculated at the corresponding UE of the multiple UEs as one of positive or negative; and update the parameter in the parameters of the machine learning model based on the aggregated analog signal obtained in the single RE. The federated ML parameter processing component 199 may be within the processor 1912. The federated ML parameter processing component 199 may be one or more hardware components specifically configured to implement the process / algorithm, implemented by one or more processors configured to execute the process / algorithm, stored in a computer-readable medium for implementation by one or more processors, or some combination thereof. The network entity 1960 may include a variety of components configured for various functions. In one configuration, the network entity 1960 includes: a component for outputting parameters of a machine learning model for multiple UEs for transmission; a component for obtaining at least one aggregated analog signal in a pair of REs designated for multiple UEs to indicate a gradient value of a parameter in the parameters of the machine learning model for the multiple UEs, the at least one aggregated analog signal obtained in the pair of REs representing a plurality of analog signals cumulatively obtained from the multiple UEs at a network node, each of the plurality of analog signals indicating a gradient value of the parameter in the parameters of the machine learning model calculated at a corresponding UE of the multiple UEs, the gradient value including a positive gradient value or a negative gradient value; and a component for updating the parameter in the parameters of the machine learning model based on the at least one aggregated analog signal obtained in the pair of REs.In one configuration, at least one aggregated analog signal includes a first aggregated analog signal and a second aggregated analog signal, and a pair of REs designated for indicating the parameter in the parameter of the machine learning model includes the first RE and the second RE; the first aggregated analog signal obtained in the first RE represents a first group of analog signals cumulatively obtained at the network node, each analog signal in the first group of analog signals indicating a positive gradient value; and the second aggregated analog signal obtained in the second RE represents a second group of analog signals cumulatively obtained at the network node, each analog signal in the second group of analog signals indicating a negative gradient value. In one configuration, the network entity 1960 also includes a component for comparing a first magnitude of the first aggregated analog signal obtained in the first RE with a second magnitude of the second aggregated analog signal obtained in the second RE, and wherein the parameter is updated based on the comparison result of the first magnitude and the second magnitude. In one configuration, the network entity 1960 also includes a component for generating an aggregated gradient of the parameter based on the comparison result of the first magnitude and the second magnitude, wherein the parameter is updated based on the aggregated gradient of the parameter. In one configuration, based on the first magnitude being greater than the second magnitude, the aggregate gradient of the parameter has a value of +p, based on the first magnitude being less than or equal to the second magnitude, the aggregate gradient of the parameter has a value of -p, and p is a real number. In one configuration, the network entity 1960 also includes a component for outputting information associated with updated parameters of the machine learning model to multiple UEs for transmission. In one configuration, the information associated with the updated parameters includes the updated parameters. In one configuration, the network entity 1960 also includes a component for outputting an indication of a pair of REs for multiple UEs for transmission, and wherein at least one aggregated analog signal is obtained in the pair of REs based on the indication of the pair of REs. In one configuration, the network entity 1960 includes: a component for outputting parameters of a machine learning model for multiple UEs for transmission; a component for obtaining an aggregated analog signal in a single RE designated for multiple UEs to indicate a sign of a gradient value relative to one of the parameters of the machine learning model for multiple UEs, the aggregated analog signal obtained in the single RE representing multiple analog signals cumulatively obtained from multiple UEs at a network node, each of the multiple analog signals indicating a sign of a gradient value relative to the parameter in the parameters of the machine learning model calculated at a corresponding UE of the multiple UEs as one of positive or negative; and a component for updating the parameter in the parameters of the machine learning model based on the aggregated analog signal obtained in the single RE. In one configuration, the presence of the analog signal in the single RE indicates that the sign of the gradient value is positive, and the absence of the analog signal in the single RE indicates that the sign of the gradient value is negative. In one configuration, the network entity 1960 also includes a component for comparing the magnitude of the aggregated analog signal obtained in the single RE with a threshold value, wherein the parameter is updated based on the comparison result of the magnitude with the threshold value.In one configuration, the network entity 1960 also includes a component for generating an aggregate gradient of the parameter based on a comparison result of the magnitude with a threshold, wherein the parameter is updated based on the aggregate gradient of the parameter. In one configuration, based on the magnitude being greater than the threshold, the aggregate gradient of the parameter has a value of +p, based on the magnitude being less than or equal to the threshold, the aggregate gradient of the parameter has a value of -p, and p is a real number. In one configuration, the network entity 1960 also includes a component for outputting information associated with updated parameters of the machine learning model to multiple UEs for transmission. In one configuration, the information associated with the updated parameters includes the updated parameters. In one configuration, the network entity 1960 also includes a component for outputting an indication of a single RE for multiple UEs for transmission, wherein at least one aggregated analog signal is obtained in the single RE based on the indication of the single RE. The component can be a federated ML parameter processing component 199 of the network entity 1960 configured to perform the functions recorded by the component.

[0200] Based on some aspects of the present disclosure, a UE receives parameters of a machine learning model from a network node; calculates a gradient value relative to one of the parameters of the machine learning model, the gradient value including a positive gradient value or a negative gradient value. The UE may send an analog signal indicating the magnitude of the gradient value of the parameter relative to the parameters of the machine learning model to the network node in one of a pair of REs. The UE may send an analog signal to the network node in a single RE based on the gradient value or skip sending the analog signal, and the single RE is designated to indicate a positive gradient value or a negative gradient value of the gradient value of the parameter relative to the parameters of the machine learning model. The network node may receive at least one aggregated analog signal in a pair of REs or a single RE, and update the parameter based on the at least one aggregated analog signal received. The network node may send information associated with the updated parameter.

[0201] It should be understood that the specific order or hierarchy of the blocks in the disclosed process / flowchart is merely an illustration of the exemplary method. It should be understood that the specific order or hierarchy of the blocks in the process / flowchart can be rearranged based on design preferences. Further, some blocks can be combined or omitted. The attached method claims provide the elements of each block in a sample order, but are not limited to the specific order or hierarchy provided.

[0202] The foregoing description is provided to enable any person skilled in the art to practice the various aspects described herein. Various modifications to these aspects will be apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects. Therefore, the claims are not limited to the various aspects described herein, but should be given the full scope consistent with the language claims. Unless otherwise specified, reference to an element in the singular form does not mean "one and only one", but "one or more". Terms such as "if", "when ......" and "while" do not mean a direct temporal relationship or reaction. That is, these phrases, such as "when ......", do not mean an immediate action in response to the occurrence of an action or during the occurrence of an action, but simply imply that if the conditions are met, the action will occur, but no specific or immediate time limit is required for the occurrence of the action. The word "exemplary" is used herein to mean "used as an example, instance, or illustration". Any aspect described as "exemplary" herein is not necessarily interpreted as being preferred or having advantages over other aspects. Unless otherwise specified, the term "some" refers to one or more. Combinations such as "at least one of A, B, or C", "one or more of A, B, or C", "at least one of A, B and C", "one or more of A, B, and C", and "A, B, C, or any combination thereof" include any combination of A, B, and / or C, and may include multiple A, multiple B, or multiple C. Specifically, combinations such as "at least one of A, B, or C", "one or more of A, B, or C", "at least one of A, B, and C", "one or more of A, B, and C", and "A, B, C, or any combination thereof" may be only A, only B, only C, A and B, A and C, B and C, or A and B and C, where any such combination may contain one or more members of A, B, or C. A set should be interpreted as a set of elements, where the number of elements is one or more. Thus, for a set of X, X will include one or more elements. If the first device receives data from the second device or sends data to the second device, data can be directly received / sent between the first device and the second device, or data can be indirectly received / sent between the first device and the second device through a collection of devices. All structural and functional equivalents of the elements throughout the various aspects described in this disclosure that are known or will be known later to those of ordinary skill in the art are expressly incorporated herein by reference and are covered by the claims. In addition, nothing disclosed herein is intended to be dedicated to the public, regardless of whether such disclosure is explicitly stated in the claims. Words such as "module", "mechanism", "element", "device" cannot replace the word "component". Therefore, no claim element will be interpreted as a functional component unless the element is explicitly recorded using the phrase "component for..."

[0203] As used herein, the phrase "based on" should not be interpreted as referring to a closed set of information, one or more conditions, one or more factors, etc. In other words, the phrase "based on A" (where "A" can be information, a condition, a factor, etc.) should be interpreted as "based at least on A" unless specifically stated differently.

[0204] The following aspects are merely illustrative and may be combined with other aspects or teachings described herein without limitation.

[0205] Aspect 1 is a method for performing wireless communication at a UE, the method comprising: receiving parameters of a machine learning model from a network node; calculating a gradient value relative to one of the parameters of the machine learning model, the gradient value comprising a positive gradient value or a negative gradient value; and sending an analog signal indicating the magnitude of the gradient value of the parameter relative to the parameters of the machine learning model to the network node in one of a pair of REs, the pair of REs being designated for indicating the gradient value relative to the parameter of the parameters of the machine learning model.

[0206] Aspect 2 is a method according to Aspect 1, wherein the pair of REs designated for indicating the parameters of the machine learning model includes: a first RE, the first RE being designated for indicating the positive gradient value; and a second RE, the second RE being designated for indicating the negative gradient value, wherein the analog signal is sent in the first RE or the second RE based on the gradient value.

[0207] Aspect 3 is a method according to any one of Aspect 1 and Aspect 2, the method further comprising: receiving information associated with updated parameters of the machine learning model from the network node, the updated parameters being at least partially based on the simulated signal sent to the network node; and updating the parameters of the machine learning model based on the information associated with the updated parameters received from the network node.

[0208] Aspect 4 is a method according to aspect 3, wherein the information associated with the updated parameters includes the updated parameters based at least in part on the simulated signal sent to the network node.

[0209] Aspect 5 is a method according to any one of Aspects 1 to 4, the method further comprising: receiving an indication of the pair of REs from the network node, wherein the analog signal is sent in the one RE in the pair of REs based on the indication of the pair of REs received from the network node.

[0210] Aspect 6 is an apparatus for wireless communication, the apparatus comprising at least one processor coupled to a memory and configured to implement any one of Aspects 1 to 5, the apparatus further comprising a transceiver coupled to the at least one processor.

[0211] Aspect 7 is an apparatus for wireless communication, the apparatus comprising means for implementing any one of Aspects 1 to 5.

[0212] Aspect 8 is a non-transitory computer-readable medium storing computer-executable code, wherein the code, when executed by a processor, causes the processor to implement any one of aspects 1 to 5.

[0213] Aspect 9 is a method for performing wireless communication at a UE, the method comprising: receiving parameters of a machine learning model from a network node; calculating a gradient value relative to one of the parameters of the machine learning model, the sign of the gradient value being one of positive or negative; and sending an analog signal to the network node in a single RE or skipping the sending of the analog signal based on the sign of the gradient value relative to the parameter in the parameters of the machine learning model, the single RE being designated for indicating the sign of the gradient value relative to the parameter in the parameters of the machine learning model.

[0214] Aspect 10 is a method according to aspect 9, wherein the presence of the analog signal in the single RE indicates that the sign of the gradient value is positive, and the absence of the analog signal in the RE indicates that the sign of the gradient value is negative.

[0215] Aspect 11 is a method according to any one of Aspect 9 and Aspect 10, the method further comprising: receiving information associated with updated parameters of the machine learning model from the network node, the updated parameters being at least partially based on the simulated signal sent to the network node; and updating the parameters of the machine learning model based on the information associated with the updated parameters received from the network node.

[0216] Aspect 12 is a method according to aspect 11, wherein the information associated with the updated parameters includes the updated parameters based at least in part on the simulated signal sent to the network node.

[0217] Aspect 13 is a method according to any one of Aspects 9 to 12, the method further comprising: receiving an indication of the single RE from the network node, wherein the analog signal is sent in the single RE based on the indication of the single RE received from the network node.

[0218] Aspect 14 is an apparatus for wireless communication, the apparatus comprising at least one processor coupled to a memory and configured to implement any one of Aspects 9 to 14, the apparatus further comprising a transceiver coupled to the at least one processor.

[0219] Aspect 15 is an apparatus for wireless communication, the apparatus comprising means for implementing any one of Aspects 9 to 14.

[0220] Aspect 16 is a non-transitory computer-readable medium storing computer-executable code, wherein the code, when executed by a processor, causes the processor to implement any one of aspects 9 to 14.

[0221] Aspect 17 is a method for performing wireless communication at a network node, the method comprising: outputting parameters of a machine learning model for multiple UEs for transmission; obtaining at least one aggregated analog signal in a pair of REs designated for the multiple UEs to indicate a gradient value of one of the parameters of the machine learning model relative to the multiple UEs, the at least one aggregated analog signal obtained in the pair of REs representing multiple analog signals cumulatively obtained from the multiple UEs at the network node, each of the multiple analog signals indicating the gradient value of the parameter in the parameters of the machine learning model calculated at a corresponding UE among the multiple UEs, the gradient value including a positive gradient value or a negative gradient value; and updating the parameter in the parameters of the machine learning model based on the at least one aggregated analog signal obtained in the pair of REs.

[0222] Aspect 18 is a method according to Aspect 17, wherein the at least one aggregated analog signal includes a first aggregated analog signal and a second aggregated analog signal, and the pair of REs designated for indicating the parameters among the parameters of the machine learning model includes a first RE and a second RE; the first aggregated analog signal obtained in the first RE represents a first group of analog signals cumulatively obtained at the network node, each analog signal in the first group of analog signals indicating the positive gradient value; and the second aggregated analog signal obtained in the second RE represents a second group of analog signals cumulatively obtained at the network node, each analog signal in the second group of analog signals indicating the negative gradient value.

[0223] Aspect 19 is a method according to Aspect 18, the method further comprising: comparing a first magnitude of the first aggregate analog signal obtained in the first RE with a second magnitude of the second aggregate analog signal obtained in the second RE, wherein the parameter is updated based on the comparison result of the first magnitude and the second magnitude.

[0224] Aspect 20 is a method according to aspect 19, the method further comprising: generating an aggregate gradient of the parameter based on the comparison result of the first magnitude and the second magnitude, wherein the parameter is updated based on the aggregate gradient of the parameter.

[0225] Aspect 21 is a method according to any one of Aspect 19 and Aspect 20, wherein based on the first value being greater than the second value, the aggregate gradient of the parameter has a value +p, based on the first value being less than or equal to the second value, the aggregate gradient of the parameter has a value -p, and p is a real number.

[0226] Aspect 22 is a method according to any one of Aspects 17 to 21, the method further comprising: outputting information associated with the updated parameters of the machine learning model to multiple UEs for sending.

[0227] Aspect 23 is a method according to aspect 22, wherein the information associated with the updated parameters includes the updated parameters.

[0228] Aspect 24 is a method according to any one of Aspects 17 to 23, the method further comprising: outputting an indication of the pair of REs for the multiple UEs for sending, wherein the at least one aggregated analog signal is obtained in the pair of REs based on the indication of the pair of REs.

[0229] Aspect 25 is an apparatus for wireless communication, the apparatus comprising at least one processor coupled to a memory and configured to implement any one of Aspects 17 to 24, the apparatus further comprising a transceiver coupled to the at least one processor.

[0230] Aspect 26 is an apparatus for wireless communication, the apparatus comprising means for implementing any one of Aspects 17 to 24.

[0231] Aspect 27 is a non-transitory computer-readable medium storing computer-executable code, wherein the code, when executed by a processor, causes the processor to implement any one of Aspects 17 to 24.

[0232] Aspect 28 is a method for performing wireless communication at a network node, the method comprising: outputting parameters of a machine learning model for multiple UEs for transmission; obtaining an aggregated analog signal in a single RE designated for the multiple UEs to indicate the sign of a gradient value of one of the parameters of the machine learning model of the multiple UEs, wherein the aggregated analog signal obtained in the single RE represents a plurality of analog signals cumulatively obtained from the multiple UEs at the network node, each of the plurality of analog signals indicating the sign of the gradient value of the parameter in the parameters of the machine learning model calculated at a corresponding UE in the multiple UEs as one of positive or negative; and updating the parameters in the parameters of the machine learning model based on the aggregated analog signal obtained in the single RE.

[0233] Aspect 29 is a method according to aspect 28, wherein the presence of the analog signal in the single RE indicates that the sign of the gradient value is positive, and the absence of the analog signal in the single RE indicates that the sign of the gradient value is negative.

[0234] Aspect 30 is a method according to any one of Aspect 28 and Aspect 29, the method further comprising: comparing the magnitude of the aggregate analog signal obtained in the single RE with a threshold, wherein the parameter is updated based on the comparison result of the magnitude with the threshold.

[0235] Aspect 31 is a method according to aspect 30, the method further comprising: generating an aggregate gradient of the parameter based on the comparison result of the magnitude and the threshold, wherein the parameter is updated based on the aggregate gradient of the parameter.

[0236] Aspect 32 is a method according to Aspect 31, wherein based on the magnitude being greater than the threshold, the aggregate gradient of the parameter has a value of +p, based on the magnitude being less than or equal to the threshold, the aggregate gradient of the parameter has a value of -p, and p is a real number.

[0237] Aspect 33 is a method according to any one of Aspects 28 to 32, the method further comprising: outputting information associated with the updated parameters of the machine learning model to multiple UEs for sending.

[0238] Aspect 34 is a method according to aspect 33, wherein the information associated with the updated parameters includes the updated parameters.

[0239] Aspect 35 is a method according to any one of Aspects 28 to 34, the method further comprising: outputting an indication of the single RE used for the multiple UEs for sending, wherein the at least one aggregate analog signal is obtained in the single RE based on the indication of the single RE.

[0240] Aspect 36 is an apparatus for wireless communication, the apparatus comprising at least one processor coupled to a memory and configured to implement any one of Aspects 28 to 35, the apparatus further comprising a transceiver coupled to the at least one processor.

[0241] Aspect 37 is an apparatus for wireless communication, the apparatus comprising means for implementing any one of Aspect 28 to Aspect 35.

[0242] Aspect 38 is a non-transitory computer-readable medium storing computer-executable code, wherein the code, when executed by a processor, causes the processor to implement any one of Aspects 28 to 35.

Claims

1. An apparatus for wireless communication at a user equipment (UE), the apparatus comprising: Memory; and at least one processor coupled to the memory and configured to: Receive machine learning model parameters from network nodes; Calculating a gradient value relative to one of the parameters of the machine learning model, the gradient value comprising a positive gradient value or a negative gradient value; and An analog signal indicating the magnitude of the gradient value relative to the parameter in the parameters of the machine learning model is sent to the network node in one of a pair of resource elements (REs), the pair of REs being designated for indicating the gradient value relative to the parameter in the parameters of the machine learning model.

2. The apparatus of claim 1 , wherein the pair of REs designated to indicate the parameters of the machine learning model comprises: a first RE, the first RE being designated to indicate the positive gradient value; and a second RE, the second RE being designated to indicate the negative gradient value, The analog signal is sent in the first RE or the second RE based on the gradient value.

3. The apparatus of claim 1 , wherein the at least one processor is further configured to: receiving, from the network node, information associated with updated parameters of the machine learning model, the updated parameters being based at least in part on the simulated signal sent to the network node; and The parameters of the machine learning model are updated based on the information associated with the updated parameters received from the network node. 4 . The apparatus of claim 3 , wherein the information associated with the updated parameters comprises the updated parameters based at least in part on the simulated signal sent to the network node.

5. The apparatus of claim 1 , wherein the at least one processor is further configured to receive an indication of the pair of REs from the network node, and Wherein the analog signal is sent in the one RE of the pair of REs based on the indication of the pair of REs received from the network node. 6 . The apparatus of claim 1 , further comprising a transceiver coupled to the at least one processor and configured to transmit the analog signal.

7. An apparatus for wireless communication at a user equipment (UE), the apparatus comprising: Memory; and at least one processor coupled to the memory and configured to: Receive machine learning model parameters from network nodes; Calculating a gradient value with respect to one of the parameters of the machine learning model, wherein the sign of the gradient value is one of positive or negative; as well as Sending a simulated signal to the network node in a single resource element (RE) or skipping the sending of the simulated signal based on the sign of the gradient value of the parameter in the parameters of the machine learning model, the single RE being designated for indicating the sign of the gradient value of the parameter in the parameters of the machine learning model.

8. The apparatus of claim 7, wherein the presence of the analog signal in the single RE indicates that the sign of the gradient value is positive, and the absence of the analog signal in the RE indicates that the sign of the gradient value is negative.

9. The apparatus of claim 7, wherein the at least one processor is further configured to: receiving, from the network node, information associated with updated parameters of the machine learning model, the updated parameters being based at least in part on the simulated signal sent to the network node; and The parameters of the machine learning model are updated based on the information associated with the updated parameters received from the network node.

10. The apparatus of claim 9, wherein the information associated with the updated parameters comprises the updated parameters based at least in part on the simulated signal sent to the network node.

11. The apparatus of claim 7, wherein the at least one processor is further configured to receive an indication of the single RE from the network node, Wherein the analog signal is sent in the single RE based on the indication of the single RE received from the network node.

12. The apparatus of claim 7, further comprising a transceiver coupled to the at least one processor and configured to transmit the analog signal.

13. An apparatus for wireless communication at a network node, the apparatus comprising: Memory; and at least one processor coupled to the memory and configured to: outputting parameters of a machine learning model for a plurality of user equipments (UEs) for transmission; obtaining, in a pair of resource elements (REs) designated for the plurality of UEs, at least one aggregated analog signal to indicate a gradient value of one of the parameters of the machine learning model relative to the plurality of UEs, the at least one aggregated analog signal obtained in the pair of REs representing a plurality of analog signals cumulatively obtained from the plurality of UEs at the network node, each of the plurality of analog signals indicating the gradient value of the parameter in the parameters of the machine learning model calculated at a corresponding UE among the plurality of UEs, the gradient value comprising a positive gradient value or a negative gradient value; and The parameters of the parameters of the machine learning model are updated based on the at least one aggregated simulated signal obtained in the pair of REs.

14. The apparatus of claim 13, wherein the at least one aggregate analog signal comprises a first aggregate analog signal and a second aggregate analog signal, and the pair of REs designated for indicating the parameter in the parameters of the machine learning model comprises a first RE and a second RE, wherein the first aggregate analog signal obtained in the first RE represents a first set of analog signals cumulatively obtained at the network node, each analog signal in the first set of analog signals indicating the positive gradient value, and The second aggregate analog signal obtained in the second RE represents a second group of analog signals cumulatively obtained at the network node, each analog signal in the second group of analog signals indicates the negative gradient value.

15. The apparatus of claim 14, wherein the at least one processor is further configured to compare a first magnitude of the first aggregate analog signal obtained in the first RE with a second magnitude of the second aggregate analog signal obtained in the second RE, and The parameter is updated based on a comparison result of the first magnitude and the second magnitude.

16. The apparatus of claim 15, wherein the at least one processor is further configured to generate an aggregate gradient of the parameter based on the comparison result of the first magnitude and the second magnitude, Wherein the parameter is updated based on the aggregated gradient of the parameter.

17. The apparatus of claim 15, wherein the aggregate gradient of the parameter has a value of +p based on the first magnitude being greater than the second magnitude, wherein the aggregate gradient of the parameter has a value -p based on the first magnitude being less than or equal to the second magnitude, and wherein p is a real number.

18. The apparatus of claim 13, wherein the at least one processor is further configured to: Information associated with the updated parameters of the machine learning model is output to multiple UEs for transmission.

19. The apparatus of claim 18, wherein the information associated with the updated parameters comprises the updated parameters.

20. The apparatus of claim 13, wherein the at least one processor is further configured to output an indication of the pair of REs for the plurality of UEs for transmission, and wherein the at least one aggregate analog signal is obtained in the pair of REs based on the indication of the pair of REs.

21. The apparatus of claim 13, further comprising a transceiver coupled to the at least one processor and configured to obtain the aggregate analog signal.

22. An apparatus for wireless communication at a network node, the apparatus comprising: Memory; and at least one processor coupled to the memory and configured to: outputting parameters of a machine learning model for a plurality of user equipments (UEs) for transmission; obtaining an aggregate analog signal in a single resource element (RE) designated for the plurality of UEs to indicate a sign of a gradient value relative to one of the parameters of the machine learning model of the plurality of UEs, the aggregate analog signal obtained in the single RE representing a plurality of analog signals cumulatively obtained at the network node from the plurality of UEs, each of the plurality of analog signals indicating that the sign of the gradient value relative to the parameter of the parameters of the machine learning model calculated at a corresponding UE among the plurality of UEs is one of positive or negative; as well as The parameters of the parameters of the machine learning model are updated based on the aggregated simulation signal obtained in the single RE.

23. The apparatus of claim 22, wherein the presence of the analog signal in the single RE indicates that the sign of the gradient value is positive, and the absence of the analog signal in the single RE indicates that the sign of the gradient value is negative.

24. The apparatus of claim 22, wherein the at least one processor is further configured to compare a magnitude of the aggregate analog signal obtained in the single RE with a threshold value, and The parameter is updated based on a comparison result of the magnitude with the threshold.

25. The apparatus of claim 24, wherein the at least one processor is further configured to generate an aggregate gradient of the parameter based on the comparison result of the magnitude with the threshold, Wherein the parameter is updated based on the aggregated gradient of the parameter.

26. The apparatus of claim 25, wherein the aggregate gradient of the parameter has a value of +p based on the magnitude being greater than the threshold, wherein the aggregate gradient of the parameter has a value -p based on the magnitude being less than or equal to the threshold, and wherein p is a real number.

27. The apparatus of claim 22, wherein the at least one processor is further configured to: Information associated with the updated parameters of the machine learning model is output to multiple UEs for transmission.

28. The apparatus of claim 27, wherein the information associated with the updated parameters comprises the updated parameters.

29. The apparatus of claim 22, wherein the at least one processor is further configured to output an indication of the single RE for the plurality of UEs for transmission, and wherein the at least one aggregate analog signal is obtained in the single RE based on the indication of the single RE.

30. The apparatus of claim 22, further comprising a transceiver coupled to the at least one processor and configured to obtain the aggregate analog signal.

Images