Quantized feedback in federated learning with randomization

Abstract

Various aspects of the present disclosure generally relate to wireless communication. In some aspects, a client device can determine feedback associated with a machine learning component based at least in part on the application of the machine learning component. Accordingly, the client device can transmit a quantized value based at least in part on the feedback. The quantized value is determined using randomization with a probability of one or more values between a plurality of quantized numbers based at least in part on respective distances of the feedback. Numerous other aspects are provided.

Description

[0001] Cross-references to related applications

[0002] This patent application claims priority to U.S. Provisional Patent Application No. 63 / 085,748, filed September 30, 2020, entitled “QUANTIZED FEEDBACK INFEDERATED LEARNING WITH RANDOMIZATION,” and U.S. Non-Provisional Patent Application No. 17 / 448,298, filed September 21, 2021, entitled “QUANTIZED FEEDBACK IN FEDERATED LEARNING WITH RANDOMIZATION,” which are hereby expressly incorporated by reference.

[0003] open field

[0004] Various aspects of this disclosure generally relate to wireless communication, and specifically to techniques and apparatus for transmitting and receiving quantized feedback in joint learning with randomization.

[0005] background

[0006] Wireless communication systems are widely deployed to provide a variety of telecommunications services such as telephone, video, data, messaging, and broadcasting. Typical wireless communication systems employ multiple access technologies that can support communication with multiple users by sharing available system resources (e.g., bandwidth, transmit power). 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, Time Division Synchronous Code Division Multiple Access (TD-SCDMA) systems, and Long Term Evolution (LTE). LTE / LTE-Advanced is an enhancement set of the Universal Mobile Telecommunications System (UMTS) mobile standard issued by the 3rd Generation Partnership Project (3GPP).

[0007] A wireless network may include one or more base stations that support communication for one or more user equipment (UEs). UEs may communicate with base stations via downlink and uplink communication. "Downlink" (or "DL") refers to the communication link from the base station to the UE, while "uplink" (or "UL") refers to the communication link from the UE to the base station.

[0008] The above multiple access technologies have been adopted in various telecommunications standards to provide a common protocol enabling different UEs to communicate at the city, country, region, and / or global levels. New Radio (NR) (which may be referred to as 5G) is an enhancement set to the LTE mobile standard issued by 3GPP. NR is designed to better support mobile broadband Internet access by using Orthogonal Frequency Division Multiplexing (OFDM) with a Cyclic Prefix (CP) (CP-OFDM) on the downlink, and CP-OFDM and / or Single Carrier Frequency Division Multiplexing (SC-FDM) (also known as Discrete Fourier Transform Extended OFDM (DFT-s-OFDM)) on the uplink, as well as supporting beamforming, multiple-input multiple-output (MIMO) antenna technologies and carrier aggregation to improve spectral efficiency, reduce costs, improve service, utilize new spectrum, and better integrate with other open standards. Further improvements to LTE, NR, and other radio access technologies remain useful as the demand for mobile broadband access continues to grow.

[0009] Overview

[0010] Some aspects described herein relate to a wireless communication method performed by a client device. The method may include determining feedback associated with an applied machine learning component, at least in part. The method may further include transmitting quantized values, at least in part in relation to the feedback, wherein the quantized values ​​are determined using randomization of probabilities having corresponding distances between one or more values, at least in part in relation to the feedback, and a plurality of quantized numbers.

[0011] Some aspects described herein relate to a wireless communication method performed by a server device. The method may include transmitting to a client device a configuration associated with a machine learning component, wherein the machine learning component accepts one or more inputs to generate one or more outputs. The method may further include receiving quantized values ​​based at least in part on feedback from the client device to which the machine learning component has been applied, wherein the quantized values ​​are based at least in part on a randomization having probabilities of corresponding distances between one or more values ​​of the feedback and a plurality of quantized numbers.

[0012] Some aspects described herein relate to an apparatus for wireless communication at a client device. The client device may include a memory and one or more processors coupled to the memory. The one or more processors may be configured to determine feedback associated with the machine learning components, at least in part, based on an applied machine learning component. The one or more processors may be further configured to transmit quantized values, at least in part, based on the feedback, wherein the quantized values ​​are determined using randomization of probabilities having corresponding distances between the one or more values, at least in part, based on the feedback, and a plurality of quantized numbers.

[0013] Some aspects described herein relate to an apparatus for wireless communication at a server device. The server device may include memory and one or more processors coupled to the memory. The one or more processors may be configured to transmit configuration associated with a machine learning component to a client device, wherein the machine learning component accepts one or more inputs to generate one or more outputs. The one or more processors may be further configured to receive quantized values ​​based at least in part on feedback from a client device to which the machine learning component has been applied, wherein the quantized values ​​are based at least in part on randomization having probabilities of corresponding distances at least in part to one or more values ​​of the feedback and a plurality of quantized numbers.

[0014] Some aspects described herein relate to a non-transient computer-readable medium storing a set of instructions for wireless communication. When executed by one or more processors of a client device, the one or more instructions cause the client device to: determine feedback associated with an applied machine learning component, at least in part, based on the application of the machine learning component. The one or more instructions, when executed by one or more processors of the client device, further cause the client device to: transmit quantized values, at least in part, based on the feedback, wherein the quantized values ​​are determined using a randomization of probabilities having corresponding distances between one or more values, at least in part, based on the feedback, and a plurality of quantized numbers.

[0015] Some aspects described herein relate to a non-transient computer-readable medium storing a set of instructions for wireless communication. When executed by one or more processors of a server device, the one or more instructions cause the server device to: transmit configuration associated with a machine learning component to a client device, wherein the machine learning component accepts one or more inputs to generate one or more outputs. The one or more instructions, when executed by one or more processors of the server device, further cause the server device to: receive quantized values ​​based at least in part on feedback from a client device to which the machine learning component has been applied, wherein the quantized values ​​are based at least in part on randomization having probabilities of corresponding distances at least in part to the feedback values ​​and a plurality of quantized numbers.

[0016] Some aspects described herein relate to a device for wireless communication. The device may include means for determining feedback associated with a machine learning component, at least in part, based on an applied machine learning component. The device may further include means for transmitting quantized values, at least in part, based on the feedback, wherein the quantized values ​​are determined using randomization of probabilities having corresponding distances between one or more values, at least in part, based on the feedback, and a plurality of quantized numbers.

[0017] Some aspects described herein relate to a device for wireless communication. The device may include means for transmitting to a client device a configuration associated with a machine learning component, wherein the machine learning component accepts one or more inputs to generate one or more outputs. The device may further include means for receiving quantized values ​​based at least in part on feedback from a client device to which the machine learning component has been applied, wherein the quantized values ​​are based at least in part on randomization having probabilities of corresponding distances between one or more values ​​of the feedback and a plurality of quantized numbers.

[0018] The aspects generally include, as substantially described herein with reference to the accompanying drawings and description, methods, apparatus, systems, computer program products, non-transient computer-readable media, user equipment, base stations, wireless communication equipment, and / or processing systems.

[0019] The foregoing has broadly outlined the features and technical advantages of the examples according to this disclosure in an effort to facilitate a better understanding of the following detailed description. Additional features and advantages will be described thereafter. The disclosed concepts and specific examples can be readily used as the basis for modifying or designing other structures for implementing the same purposes as this disclosure. Such equivalent constructions do not depart from the scope of the appended claims. The characteristics of the concepts disclosed herein, in both their organization and manner of operation, and their associated advantages, will be better understood by considering the following description in conjunction with the accompanying drawings. Each drawing is provided for illustrative and descriptive purposes and not for defining limitations on the claims.

[0020] While aspects are described herein by way of example, those skilled in the art will understand that such aspects can be implemented in many different arrangements and scenarios. The techniques described herein can be implemented using different platform types, devices, systems, shapes, sizes, and / or package layouts. For example, some aspects may be implemented via integrated chip embodiments or other devices based on non-modular components (e.g., end-user devices, vehicles, communication devices, computing devices, industrial equipment, retail / shopping devices, medical devices, and / or artificial intelligence devices). Aspects can be implemented in chip-level components, modular components, non-modular components, non-chip-level components, device-level components, and / or system-level components. Devices incorporating the described aspects and features may include additional components and features for implementing and practicing the claimed and described aspects. For example, the transmission and reception of wireless signals may include one or more components for analog and digital purposes (e.g., hardware components including antennas, radio frequency (RF) chains, power amplifiers, modulators, buffers, processors, interleavers, adders, and / or summers). The aspects described herein are intended to be practiced in a wide variety of devices, components, systems, distributed arrangements, and / or end-user devices of various sizes, shapes, and configurations. Brief description of the attached diagram

[0022] To gain a more detailed understanding of the features described above in this disclosure, reference can be made to various aspects of the above brief overview, some of which are illustrated in the accompanying drawings. However, it should be noted that the drawings illustrate only certain typical aspects of this disclosure and should not be considered as limiting its scope, as other equivalent aspects are permissible in this description. Identical reference numerals in different drawings may identify the same or similar elements.

[0023] Figure 1 This is a diagram illustrating an example of a wireless network according to this disclosure.

[0024] Figure 2 This is a diagram illustrating an example of communication between a base station and a user equipment (UE) in a wireless network according to this disclosure.

[0025] Figure 3 This is a diagram illustrating an example of joint learning for machine learning components according to this disclosure.

[0026] Figure 4 This is a diagram illustrating an example of transmitting and receiving quantized feedback in joint learning with randomization, according to the present disclosure.

[0027] Figure 5 and 6 This is a diagram illustrating an example process associated with transmitting and receiving quantized feedback in joint learning with randomization, according to the present disclosure.

[0028] Figure 7 and 8 This is a diagram illustrating an example apparatus for wireless communication based on this disclosure.

[0029] Detailed description

[0030] The various aspects of this disclosure are described more fully below with reference to the accompanying drawings. However, this disclosure may be implemented in many different forms and should not be construed as being limited to any specific structure or function given throughout this disclosure. Rather, these aspects are provided so that this disclosure will be thorough and complete, and will fully convey the scope of this disclosure to those skilled in the art. Those skilled in the art will appreciate that the scope of this disclosure is intended to cover any aspect of this disclosure disclosed herein, whether implemented independently of or in combination with any other aspect of this disclosure. For example, any number of aspects set forth herein may be used to implement an apparatus or practice. Furthermore, the scope of this disclosure is intended to cover such apparatuses or methods practiced using additional structures, functionalities, or structures and functionalities that complement or supplement the various aspects of this disclosure set forth herein. It should be understood that any aspect of this disclosure disclosed herein may be implemented by one or more elements of the claims.

[0031] Several aspects of a telecommunications system will now be described with reference to various devices and techniques. These devices and techniques will be described in the following detailed description and explained in the accompanying drawings by various blocks, modules, components, circuits, steps, processes, algorithms, etc. (collectively, "elements"). These elements can be implemented using hardware, software, or a combination thereof. Whether such elements are implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system.

[0032] While the aspects herein may be described using terms commonly associated with 5G or New Radio (NR) Radio Access Technology (RAT), the aspects of this disclosure may be applied to other RATs, such as 3G RAT, 4G RAT, and / or RATs after 5G (e.g., 6G).

[0033] Figure 1 This is a diagram illustrating an example of a wireless network 100 according to this disclosure. The wireless network 100 may be a 5G (e.g., NR) network and / or a 4G (e.g., Long Term Evolution (LTE) network) network, or may include elements thereof. The wireless network 100 may include one or more base stations 110 (shown as BS 110a, BS 110b, BS 110c, and BS 110d), one or more user equipment (UE) 120 (shown as UE 120a, UE 120b, UE 120c, UE 120d, and UE 120e), and / or other network entities. Base station 110 is the entity that communicates with UE 120. Base station 110 (sometimes referred to as BS) may include, for example, an NR base station, an LTE base station, a B-node, an eNB (e.g., in 4G), a gNB (e.g., in 5G), an access point, and / or a transmit / receive point (TRP). Each base station 110 may provide communication coverage for a specific geographic area. In the 3rd Generation Partnership Project (3GPP), the term "cell" can refer to the coverage area of ​​base station 110 and / or the base station subsystem serving that coverage area, depending on the context in which the term is used.

[0034] Base station 110 provides communication coverage to macrocells, picocells, femtocells, and / or another type of cell. Macrocells can cover a relatively large geographic area (e.g., a radius of several kilometers) and allow unrestricted access by UE 120 with a service subscription. Picocells can cover a relatively small geographic area and allow unrestricted access by UE 120 with a service subscription. Femtocells can cover a relatively small geographic area (e.g., a residential area) and allow restricted access by UE 120 associated with that femtocell (e.g., UE 120 in a closed subscriber group (CSG)). Base station 110 for macrocells may be referred to as a macro base station. Base station 110 for picocells may be referred to as a pico base station. Base station 110 for femtocells may be referred to as a femtocell or a home base station. Figure 1 In the example shown, BS 110a can be a macro base station for macro cell 102a, BS 110b can be a pico base station for pico cell 102b, and BS 110c can be a femto base station (BS) for femtocell 102c. A base station may support one or more (e.g., three) cells.

[0035] In some examples, the cell may not necessarily be stationary, and the geographical area of ​​the cell may move depending on the location of the mobile base station 110 (e.g., a mobile base station). In some examples, base stations 110 may interconnect with each other and / or interconnect to one or more other base stations 110 or network nodes (not shown) in the wireless network 100 using any suitable transport network via various types of backhaul interfaces (such as direct physical connections or virtual networks).

[0036] Wireless network 100 may include one or more relay stations. A relay station is an entity capable of receiving data transmissions from an upstream station (e.g., base station 110 or UE 120) and transmitting those data transmissions to a downstream station (e.g., UE 120 or base station 110). A relay station may be a UE 120 capable of relaying transmissions for other UE 120s. Figure 1 In the example shown, BS 110d (e.g., a relay base station) can communicate with BS 110a (e.g., a macro base station) and UE 120d to facilitate communication between BS 110a and UE 120d. The base station 110 for relay communication may be referred to as a relay station, relay base station, relay, etc.

[0037] Wireless network 100 can be a heterogeneous network comprising different types of base stations 110 (such as macro base stations, pico base stations, femto base stations, or relay base stations, etc.). These different types of base stations 110 may have different transmit power levels, different coverage areas, and / or different effects on interference in wireless network 100. For example, macro base stations may have high transmit power levels (e.g., 5 to 40 watts), while pico base stations, femto base stations, and relay base stations may have lower transmit power levels (e.g., 0.1 to 2 watts).

[0038] Network controller 130 can be coupled to or communicate with a group of base stations 110 and can provide coordination and control over these base stations 110. Network controller 130 can communicate with base stations 110 via backhaul communication links. Base stations 110 can communicate with each other directly or indirectly via wireless or wired backhaul communication links.

[0039] Each UE 120 may be distributed throughout the wireless network 100, and each UE 120 may be stationary or mobile. UE 120 may include, for example, access terminals, terminals, mobile stations, and / or subscriber units. UE 120 may be a cellular phone (e.g., a smartphone), a personal digital assistant (PDA), a wireless modem, a wireless communication device, a handheld device, a laptop computer, a cordless phone, a wireless local loop (WLL) station, a tablet device, a camera, a gaming device, a netbook, a smartbook, an ultrabook, a medical device, a biometric device, a wearable device (e.g., a smartwatch, smart clothing, smart glasses, a smart wristband, smart jewelry (e.g., a smart ring or smart bracelet)), an entertainment device (e.g., a music device, a video device, and / or a satellite radio), an in-vehicle component or sensor, a smart meter / sensor, industrial manufacturing equipment, a GPS device, or any other suitable device configured to communicate via a wireless medium.

[0040] Some UEs 120 may be considered Machine-Type Communication (MTC) UEs, or evolved or enhanced Machine-Type Communication (eMTC) UEs. MTC UEs and / or eMTC UEs may include, for example, robots, drones, remote devices, sensors, meters, monitors, and / or location tags, which can communicate with a base station, another device (e.g., a remote device), or some other entity. Some UEs 120 may be considered Internet of Things (IoT) devices, and / or may be implemented as NB-IoT (Narrowband IoT) devices. Some UEs 120 may be considered client equipment. UE 120 may be included within a housing that houses the components of UE 120, such as processor components and / or memory components. In some examples, the processor components and memory components may be coupled together. For example, the processor components (e.g., one or more processors) and memory components (e.g., memory) may be operatively coupled, communicatively coupled, electronically coupled, and / or electrically coupled.

[0041] Generally, any number of wireless networks 100 can be deployed in a given geographical area. Each wireless network 100 can support a specific RAT and can operate on one or more frequencies. A RAT may be referred to as a radio technology, air interface, etc. A frequency may be referred to as a carrier, frequency channel, etc. Each frequency can support a single RAT in a given geographical area to avoid interference between wireless networks using different RATs. In some cases, NR or 5G RAT networks can be deployed.

[0042] In some examples, two or more UEs 120 (e.g., shown as UE 120a and UE 120e) may communicate directly using one or more sidelink channels (e.g., without using base station 110 as an intermediary). For example, UEs 120 may communicate using peer-to-peer (P2P) communication, device-to-device (D2D) communication, vehicle-to-everything (V2X) protocols (e.g., which may include vehicle-to-vehicle (V2V) protocols, vehicle-to-infrastructure (V2I) protocols, or vehicle-to-pedestrian (V2P) protocols), and / or mesh networks. In such examples, UEs 120 may perform scheduling operations, resource selection operations, and / or other operations described elsewhere herein as performed by base station 110.

[0043] Devices in Wireless Network 100 can communicate using the electromagnetic spectrum, which can be subdivided into various categories, bands, channels, etc., according to frequency or wavelength. For example, each device in Wireless Network 100 can communicate using one or more operating frequency bands. In 5G NR, two initial operating frequency bands have been designated as frequency ranges FR1 (410MHz–7.125GHz) and FR2 (24.25GHz–52.6GHz). It should be understood that although a portion of FR1 is greater than 6GHz, FR1 is generally (interchangeably) referred to as the “sub-6GHz” band in various documents and articles. Similar naming issues sometimes arise regarding FR2; although different from the Very High Frequency (EHF) band (30GHz–300GHz) designated as the “millimeter wave” band by the International Telecommunication Union (ITU), FR2 is generally (interchangeably) referred to as the “millimeter wave” band in various documents and articles.

[0044] The frequencies between FR1 and FR2 are generally referred to as intermediate frequency (IF) bands. Recent 5G NR studies have designated the operating bands of these IF bands as the frequency range designation FR3 (7.125 GHz – 24.25 GHz). Bands falling within FR3 can inherit FR1 and / or FR2 characteristics, thus effectively extending the features of FR1 and / or FR2 into the IF band. Additionally, higher frequency bands are currently being explored to extend 5G NR operation above 52.6 GHz. For example, three higher operating frequency bands have been designated as the frequency range designations FR4a or FR4-1 (52.6 GHz – 71 GHz), FR4 (52.6 GHz – 114.25 GHz), and FR5 (114.25 GHz – 300 GHz). Each of these higher frequency bands falls within the EHF band.

[0045] Considering the examples above, unless otherwise stated, it should be understood that, as used herein, the term "sub-6GHz," etc., can broadly refer to frequencies less than 6GHz, within FR1, or that may include intermediate frequency band frequencies. Furthermore, unless otherwise stated, it should be understood that, as used herein, the term "millimeter wave," etc., can broadly refer to frequencies that may include intermediate frequency band frequencies, within FR2, FR4, FR4-a, or FR4-1 and / or FR5, or within the EHF band. It is conceivable that the frequencies included in these operating frequency bands (e.g., FR1, FR2, FR3, FR4, FR4-a, FR4-1, and / or FR5) can be modified, and the techniques described herein are applicable to those modified frequency ranges.

[0046] As indicated above, Figure 1 This is provided as an example. Other examples may differ from the one provided. Figure 1 The example described.

[0047] Figure 2 This is a diagram illustrating an example 200 of communication between a base station 110 and a UE 120 in a wireless network 100 according to this disclosure. The base station 110 may be equipped with a set of antennas 234a to 234t, such as T antennas (T≥1). The UE 120 may be equipped with a set of antennas 252a to 252r, such as R antennas (R≥1).

[0048] At base station 110, transmit processor 220 can receive data from data source 212 intended for UE 120 (or a group of UEs 120). Transmit processor 220 can select one or more modulation and coding schemes (MCS) for UE 120 based at least in part on one or more channel quality indicators (CQIs) received from UE 120. Base station 110 can process (e.g., encode and modulate) the data for UE 120 based at least in part on the MCS(s) selected for UE 120 and can provide data symbols to UE 120. Transmit processor 220 can process system information (e.g., semi-static resource partitioning information (SRPI)) and control information (e.g., CQI requests, grants, and / or higher-layer signaling) and provide overhead symbols and control symbols. Transmit processor 220 can generate reference symbols for reference signals (e.g., a cell-specific reference signal (CRS) or demodulation reference signal (DMRS)) and synchronization signals (e.g., a primary synchronization signal (PSS) or secondary synchronization signal (SSS)). Transmit (TX) multiple-input multiple-output (MIMO) processor 230 can perform spatial processing (e.g., precoding) on ​​data symbols, control symbols, overhead symbols, and / or reference symbols where applicable, and can provide a set of output symbol streams (e.g., T output symbol streams) to a corresponding set of modems 232 (e.g., T modulators) (shown as modems 232a to 232t). For example, each output symbol stream can be provided to a modulator component (shown as MOD) of modem 232. Each modem 232 can use a corresponding modulator component to process the corresponding output symbol stream (e.g., for OFDM) to obtain an output sample stream. Each modem 232 may further use a corresponding modulator component to process (e.g., convert to analog, amplify, filter, and / or upconvert) the output sample stream to obtain a downlink signal. Modems 232a to 232t may transmit a set of downlink signals (e.g., T downlink signals) via a corresponding set of antennas 234 (e.g., T antennas) (shown as antennas 234a to 234t).

[0049] At UE 120, an antenna set 252 (shown as antennas 252a to 252r) can receive downlink signals from base station 110 and / or other base stations 110 and can provide a set of received signals (e.g., R received signals) to a set of modems 254 (e.g., R modems) (shown as modems 254a to 254r). For example, each received signal can be provided to a demodulator component (shown as DEMOD) of modem 254. Each modem 254 can use a corresponding demodulator component to condition (e.g., filter, amplify, downconvert, and / or digitize) the received signal to obtain an input sample. Each modem 254 can use the demodulator component to further process the input sample (e.g., for OFDM) to obtain received symbols. A MIMO detector 256 can obtain the received symbols from modem 254, perform MIMO detection on these received symbols where applicable, and can provide detected symbols. The receiver processor 258 can process (e.g., demodulate and decode) these detected symbols, provide decoded data for UE 120 to data sink 260, and provide decoded control and system information to controller / processor 280. The term "controller / processor" can refer to one or more controllers, one or more processors, or a combination thereof. The channel processor can determine Reference Signal Received Power (RSRP) parameters, Received Signal Strength Indicator (RSSI) parameters, Reference Signal Received Quality (RSRQ) parameters, and / or CQI parameters, etc. In some examples, one or more components of UE 120 may be included in housing 284.

[0050] Network controller 130 may include communication unit 294, controller / processor 290, and memory 292. Network controller 130 may include one or more devices, such as those in the core network. Network controller 130 may communicate with base station 110 via communication unit 294.

[0051] One or more antennas (e.g., antennas 234a to 234t and / or antennas 252a to 252r) may include one or more antenna panels, one or more antenna groups, one or more antenna element assemblies, and / or one or more antenna arrays, etc., or may be included therein. Antenna panels, antenna groups, antenna element assemblies, and / or antenna arrays may include one or more antenna elements (within a single housing or multiple housings), coplanar antenna element assemblies, non-coplanar antenna element assemblies, and / or coupled to one or more transmit and / or receive components (such as...) Figure 2 One or more antenna elements (one or more components).

[0052] On the uplink, at UE 120, transmit processor 264 can receive and process data from data source 262 and control information from controller / processor 280 (e.g., reports including RSRP, RSSI, RSRQ, and / or CQI). Transmit processor 264 can generate reference symbols for one or more reference signals. Symbols from transmit processor 264 may be pre-encoded by TX MIMO processor 266 where applicable, further processed by modem 254 (e.g., for DFT-s-OFDM or CP-OFDM), and transmitted to base station 110. In some examples, modem 254 of UE 120 may include modulator and demodulator. In some examples, UE 120 includes a transceiver. The transceiver may include any combination of antennas 252, modems 254, MIMO detector 256, receive processor 258, transmit processor 264, and / or TX MIMO processor 266. The transceiver can be used by a processor (e.g., controller / processor 280) and memory 282 to perform aspects of any of the methods described herein (e.g., references). Figures 4 to 8 ).

[0053] At base station 110, uplink signals from UE 120 and / or other UEs may be received by antenna 234, processed by modem 232 (e.g., demodulator component of modem 232, shown as DEMOD), detected by MIMO detector 236 where applicable, and further processed by receiver processor 238 to obtain decoded data and control information transmitted by UE 120. Receiver processor 238 may provide the decoded data to data sink 239 and the decoded control information to controller / processor 240. Base station 110 may include communication unit 244 and may communicate with network controller 130 via communication unit 244. Base station 110 may include scheduler 246 to schedule one or more UEs 120 for downlink and / or uplink communication. In some examples, modem 232 of base station 110 may include modulator and demodulator. In some examples, base station 110 includes transceiver. The transceiver may include any combination of antennas 234, modems 232, MIMO detectors 236, receiver processors 238, transmitter processors 220, and / or TX MIMO processors 230. The transceiver may be used by a processor (e.g., controller / processor 240) and memory 242 to perform aspects of any of the methods described herein (e.g., references). Figures 4 to 8 ).

[0054] The controller / processor 240 of base station 110, the controller / processor 280 of UE 120, and / or Figure 2Any other component may perform one or more techniques associated with transmitting and receiving quantized feedback in joint learning with randomization, as described in more detail elsewhere herein. For example, the controller / processor 240 of base station 110, the controller / processor 280 of UE 120, and / or Figure 2 Any other component may execute or direct, for example Figure 5 Process 500 Figure 6 The operation of process 600 and / or other processes as described herein. Memory 242 and memory 282 may store data and program code for base station 110 and UE 120, respectively. In some examples, memory 242 and / or memory 282 may include a non-transitory computer-readable medium storing one or more instructions (e.g., code and / or program code) for wireless communication. For example, when executed by one or more processors of base station 110 and / or UE 120 (e.g., directly executed, or executed after compilation, transformation, and / or interpretation), the one or more processors, UE 120, and / or base station 110 may cause the one or more processors, UE 120, and / or base station 110 to perform or direct, for example... Figure 5 Process 500 Figure 6 The operation of process 600, and / or other processes described herein. In some examples, the execution instructions may include run instructions, translate instructions, compile instructions, and / or interpret instructions, etc. In some aspects, the server device described herein is base station 110, is included in base station 110, or includes... Figure 2 One or more components of the base station 110 shown. In some aspects, the client device described herein is UE 120, is included in UE 120, or includes... Figure 2 One or more components of the UE 120 shown.

[0055] In some aspects, client devices (e.g., UE 120, Figure 7 The equipment 700 and / or another client device (such as a tablet, laptop, or desktop computer) may include: means for determining feedback associated with a machine learning component based at least in part on the application of the machine learning component; and / or means for transmitting quantized values ​​based at least in part on the feedback, wherein the quantized values ​​are determined using randomization having probabilities of corresponding distances between one or more values ​​based at least in part on the feedback and a plurality of quantized numbers. In some aspects, means for enabling the client device to perform the operations described herein may include, for example, one or more of the following: antenna 252, demodulator 254, MIMO detector 256, receiver processor 258, transmitter processor 264, TX MIMO processor 266, controller / processor 280, or memory 282.

[0056] In some aspects, server equipment (e.g., base station 110, Figure 8 The equipment 800, and / or another server equipment, such as one or more server computers in a server cluster and / or at least a portion of the core network supporting base station 110, may include: for sending commands to client devices (e.g., UE 120, ...). Figure 7 The means for transmitting configurations associated with a machine learning component to an equipment 700 and / or another client device (such as a tablet, laptop, or desktop computer), wherein the machine learning component accepts one or more inputs to generate one or more outputs; and / or means for receiving quantized values ​​based at least in part on feedback from a client device to which the machine learning component has been applied, wherein the quantized values ​​are based at least in part on randomization having probabilities of corresponding distances between one or more values ​​of the feedback and a plurality of quantized numbers. In some aspects, means for a server device to perform the operations described herein may include one or more of, for example, a transmit processor 220, a TX MIMO processor 230, a modem 232, an antenna 234, a MIMO detector 236, a receive processor 238, a controller / processor 240, a memory 242, or a scheduler 246.

[0057] although Figure 2 The boxes in the diagram are interpreted as different components, but the functions described above with respect to these boxes can be implemented using a single hardware component, software component, or combination of components. For example, the functions described with respect to transmit processor 264, receive processor 258, and / or TX MIMO processor 266 can be performed by controller / processor 280 or under the control of controller / processor 280.

[0058] As indicated above, Figure 2 This is provided as an example. Other examples may differ from the one provided. Figure 2 The example described.

[0059] Figure 3 This is a diagram illustrating example 300 of joint learning for a machine learning component according to this disclosure. As shown, base station 110 can communicate with a group of UEs 120 (denoted as "UE 1,...,UE k,..., and UE K"). Base station 110 and UEs 120 can communicate via a wireless network (e.g., Figure 1 The wireless networks 100 shown communicate with each other. In some respects, any number of additional UEs 120 may be included in a set containing K UEs 120.

[0060] A machine learning component is a component (e.g., hardware, software, or a combination thereof) of a device (e.g., a client device, server device, UE, base station) that performs one or more machine learning procedures. A machine learning component may include, for example, hardware and / or software that can learn to perform procedures without being explicitly trained to do so. A machine learning component may include, for example, feature learning processing blocks and / or representation learning processing blocks. A machine learning component may include one or more neural networks. A neural network may include, for example, an autoencoder.

[0061] As shown in Example 300, federated learning can be used to train machine learning components. Federated learning is a machine learning technique that enables multiple clients (e.g., UE 120) to collaboratively train a machine learning model based on training data, without the server device (e.g., base station 110) collecting that training data from the client devices. Federated learning techniques may involve one or more global neural network models (e.g., as described in further detail below) trained from data stored on multiple client devices.

[0062] As indicated by reference numeral 310, base station 110 can transmit machine learning components to UE 120. As shown, UE 120 may each include a first communication manager 320. The first communication manager 320 may be configured to utilize the machine learning components to perform one or more wireless communication tasks and / or one or more user interface tasks. The first communication manager 320 may be configured to utilize any number of additional machine learning components.

[0063] like Figure 3 As shown, base station 110 may include a second communication manager 330. The second communication manager 330 may be configured to utilize global machine learning components to perform one or more wireless communication tasks, perform one or more user interface tasks, and / or facilitate joint learning associated with the machine learning components.

[0064] Each UE 120 can train its machine learning component locally using the training data collected by the UE 120. Each UE 120 can optimize the set of model parameters associated with the machine learning component (e.g., by w) (n) The UE 120 set can be configured to provide updates to the base station 110 multiple times (e.g., periodically, on demand, and / or while updating local machine learning components).

[0065] A “joint learning round” refers to a training session completed by UE 120 that corresponds to (e.g., precedes) the updates provided by UE 120 to base station 110. In some aspects, a joint learning round may include the transmission of updates by UE 120 and the reception of updates by base station 110. The joint learning round index (e.g., denoted by n) indicates the number of rounds since the most recent global update was transmitted from base station 110 to UE 120. Initial provisioning of machine learning components on UE 120 and / or transmitting global updates to machine learning components to UE 120, etc., can trigger the start of a new round of joint learning.

[0066] In some aspects, for example, the first communication manager 320 of UE 120 can determine updates corresponding to a machine learning component by training that machine learning component. Updates may include any updated information determined at least in part based on a training protocol associated with the machine learning component. Updates may include, for example, an updated machine learning component (e.g., an updated neural network model), an updated set of parameters (e.g., an updated set of weights for the neural network), a set of gradients associated with the loss function of the machine learning component, and / or compressed updates, etc. In some aspects, as indicated by reference numeral 340, each of UE 120 may collect training data and store that training data in a memory device. The stored training data may be referred to as a “local dataset.” As indicated by reference numeral 350, each of UE 120 may determine local updates associated with the machine learning component.

[0067] In some respects, for example, the first communication manager 320 can access training data from a memory device and use the training data to determine the input vector to be fed into the machine learning component (e.g., by x). j (represented) to generate the training output from this machine learning component (e.g., by y) j (Representation). Input vector x j It can include an array of input values ​​and the training output y. j It can include a value (e.g., a value between 0 and 9).

[0068] Training output y j The model parameters w can be used to help determine the function that maximizes the variational lower bound. (n) The negative variational lower bound function (i.e., the negative of the variational lower bound function) can correspond to the local loss function (e.g., by F...). k (w) represents that it can be represented in a form similar to the following:

[0069]

[0070] Where D kThis represents the size of the local dataset associated with UE k. The model parameters w can be optimized using the stochastic gradient descent (SGD) algorithm. (n) The first communication manager 320 of UE 120 can execute one or more SGD procedures to determine the optimization parameters w. (n) And the gradient of the loss function F(w) can be determined (e.g., denoted as...). The first communication manager 320 can further refine the machine learning component, at least in part, based on the loss function value and / or gradient, etc.

[0071] The gradient is determined by repeatedly training the machine learning component. In this process, the first communication manager 320 can determine the update corresponding to the machine learning component. Each repetition of the above training procedure can be referred to as an epoch. In some aspects, this update may include an updated set of model parameters w. (n) The updated model parameter set w (n) And the previous model parameter set w (n-1) Differences between them, gradient sets And / or updated machine learning components (e.g., updated neural network models), etc.

[0072] As shown by reference numeral 360 in the attached figure, UE 120 can transmit its respective local updates (denoted as "local update 1, ..., local update k, ..., local update K"). In some aspects, local updates may include compressed versions of local updates. For example, in some aspects, UE 120 can transmit compressed gradient sets (e.g., by...). (represented by q), where q represents the value applied to the gradient set. Compression scheme.

[0073] As shown by reference numeral 370 in the attached figure, base station 110 (e.g., using a second communication manager 330) can aggregate updates received from UE 120. For example, the second communication manager 330 can average the received gradients to determine the aggregated updates, which can be represented in a form similar to the following:

[0074]

[0075] Where, as explained above, K represents the total number of UEs 120 from which updates are received. In some examples, the second communication manager 330 may use other aggregation techniques to aggregate the received updates. As indicated by reference numeral 380, the second communication manager 330 may update the global machine learning components based on the aggregated updates. In some aspects, for example, the second communication manager 330 may aggregate the updates by grouping each dataset size (e.g., denoted as D) into smaller datasets. kThe local dataset is normalized to be treated as equal, thereby updating the global machine learning component. The second communication manager 330 can use multiple rounds of updates from UE 120 to update the global machine learning component until the global loss function is minimized. The global loss function can be given, for example, in a form similar to the following:

[0076]

[0077] Where D k =D, where D represents a normalization constant. In some respects, base station 110 may transmit updates associated with the updated global machine learning components to UE 120.

[0078] UE 120 can use machine learning components for any number of different types of operations, transmissions, and / or user experience enhancements, etc. In some aspects, UE 120 can use one or more machine learning components to report to the base station information associated with received signals, user interactions with UE 120, and / or location information. In some aspects, UE 120 can perform measurements associated with reference signals and use one or more machine learning components to facilitate reporting these measurements to the base station. For example, UE 120 can measure reference signals during beam management to achieve channel state feedback (CSF), measure the received power of reference signals from serving cells and / or neighboring cells, measure network signal strength between radio access technologies (e.g., WiFi), and / or measure sensor signals used to detect the location of one or more objects in the environment, etc. In some aspects, UE 120 can use one or more machine learning components to use data associated with user and UE 120 interactions to customize or otherwise enhance the user experience of the user interface.

[0079] Information exchange in this type of federated learning typically takes place over a Wi-Fi connection, where wired connections associated with modems, routers, and other hardware eliminate concerns about limited and / or expensive communication resources. However, implementing federated learning for machine learning components in a cellular environment can have a positive impact on network performance and user experience.

[0080] In some cases, during joint learning, the UE may incur significant network overhead and power when transmitting updates to the base station. Accordingly, to reduce overhead, the UE can quantize the update and transmit the quantized update to the base station. This quantization can be applied to a vector (e.g., the gradient as described above). ) and / or scalars (e.g., updated weights for machine learning components). However, quantization typically introduces significant errors that increase with the number of UEs used for joint learning. For example, if multiple UEs compute a scalar of 0.9 as an update, the base station might receive multiple updates from these UEs that are quantized to a scalar of 1.0, making it impossible for the base station to determine that the update was computed as 0.9 at those UEs.

[0081] Some of the techniques and apparatus described herein provide more accurate quantization of updates for joint machine learning of learning components. In some aspects, a client device (e.g., UE 120) can use a machine learning component from a server device (e.g., base station 110) to determine feedback. For example, UE 120 can locally train the machine learning component to determine local updates associated with it. UE 120 can quantize the feedback using randomization with probabilities at least partially based on the corresponding distances between one or more values ​​of the feedback and multiple quantized numbers. Accordingly, base station 110 more accurately aggregates feedback from multiple UEs, including UE 120. For example, if multiple UEs compute a scalar of 0.9 as an update, base station 110 receives multiple updates with a quantized scalar of 1.0 distributed as 90% and a quantized scalar of 0.0 as 10%, such that base station 110 can determine that the update is computed as 0.9 at the UE. As a result, network performance is improved by using quantization during joint learning without incurring a significant loss of accuracy during joint learning.

[0082] As indicated above, Figure 3 This is provided merely as an example. Other examples may differ from those provided. Figure 3 The example described.

[0083] Figure 4 This is a diagram illustrating Example 400 of transmitting and receiving quantized feedback in joint learning with randomization according to this disclosure. In Example 400, UE 120 and base station 110 can communicate with each other. In some aspects, UE 120 and base station 110 can use a wireless network (such as...) Figure 1 The UE 120 communicates with the wireless network 110. Although the description herein focuses on the UE 120 and the base station 110, the description similarly applies to other client devices (such as tablets, laptops, desktop computers, and / or other mobile or quasi-mobile devices used for federated learning) and / or other server devices (such as one or more server computers on a server cluster and / or at least a portion of the core network supporting the base station 110). Although the description herein focuses on a single UE 120, the description similarly applies to multiple UEs (e.g., as described above). Figure 3 (UE120).

[0084] As shown in conjunction with reference numeral 405, base station 110 can transmit and UE 120 can receive a configuration associated with a machine learning component, wherein the machine learning component accepts one or more inputs to generate one or more outputs. In some aspects, this configuration may be a federated learning configuration. For example, the configuration may be carried in a radio resource control (RRC) message. The configuration may indicate a machine learning component including, for example, at least one neural network model.

[0085] As shown in conjunction with reference numeral 410, UE 120 can determine the feedback associated with the machine learning component, at least in part, based on the application of the machine learning component. For example, as in conjunction with Figure 3 As described, UE 120 can access training data (e.g., stored in UE 120's memory, stored in a database accessible to UE 120, and / or received from base station 110) and use the training data to determine an input vector (e.g., denoted as xj) to be input into a machine learning component to generate a training output (e.g., denoted as yj) from the machine learning component. UE 120 can further use a local loss function (e.g., denoted as F). k (w)) to determine the feedback based at least in part on the training output yj.

[0086] In some respects, the feedback may include at least one scalar. For example, the feedback may include one or more updated weights for the machine learning component (e.g., associated with one or more nodes of at least one neural network model and / or associated with one or more nodes of at least one decision tree).

[0087] Additionally or alternatively, the feedback may include at least one vector. For example, as in combination Figure 3 As described, the feedback may include one or more gradients of the loss function F(w) (e.g., denoted as...). (For example, the SGD algorithm is used to determine the optimal model parameters w) (n) ).

[0088] As further illustrated with reference to reference numeral 410 in the accompanying drawings, UE 120 may determine the quantized value at least partially based on the feedback using randomization having probabilities based at least in part on the corresponding distances between one or more values ​​of the feedback and the plurality of quantized numbers. For example, UE 120 may quantize the feedback to encode it using fewer bits, which reduces the network overhead of transmitting the feedback to base station 110 and the memory overhead of storing the feedback at UE 120 and base station 110. UE 120 may use randomization such that base station 110 can recover additional information about the feedback from the distribution associated with feedback from multiple UEs.

[0089] In one example, UE 120 can select -3, -1, 1, and 3 as quantized numbers so that scalar feedback can be encoded using only two bits. Accordingly, UE 120 can quantize a feedback value of 0.6 to 1 with a probability of 0.95 (or 95%) based at least in part on a distance of 0.4 between the feedback value and the quantized number 1, and quantize it to -1 with a probability of 0.05 (or 5%) based at least in part on a distance of 1.6 between the feedback value and the quantized number -1. In some aspects, UE 120 can use non-uniform quantization with randomization. For example, UE 120 can select -4, -1, 1, and 4 as quantized numbers so that scalar feedback can be encoded using only two bits. Accordingly, UE 120 may quantize a feedback value of 2.0 as 1 with a probability of 0.75 (or 75%) based at least in part on a distance of 1.0 between the feedback value and the quantized digit 1, and may quantize it as 4 with a probability of 0.25 (or 25%) based at least in part on a distance of 2.0 between the feedback value and the quantized digit 4.

[0090] In some aspects, the feedback may include multiple scalars. Accordingly, the quantized value may be based at least in part on all or some of these multiple scalars. For example, when the feedback includes multiple updated weights, UE 120 may quantize the updated weights, more than one but not all of the updated weights, or one of all the updated weights. In some aspects, UE 120 may use the same quantization number and / or the same relationship between distance and probability for two or more of the multiple scalars. Additionally or alternatively, UE 120 may use different quantization numbers and / or different relationships between distance and probability for two or more of the multiple scalars.

[0091] In some aspects, the feedback may include at least one vector. Accordingly, the quantized value may be at least partially based on one component of the at least one vector. Alternatively, the quantized value may be at least partially based on two or more components of the at least one vector. For example, UE 120 may quantize some or all of the components of the at least one vector. In some aspects, UE 120 may use the same quantized number and / or the same relationship between these distances and probabilities for two or more of these components. Additionally or alternatively, UE 120 may use different quantized numbers and / or different relationships between distances and probabilities for two or more of these components.

[0092] In some aspects, the quantized value may be at least partially based on the projection of the at least one vector. For example, when the feedback includes at least one gradient (e.g., as described above), the UE 120 may project the at least one gradient along one or more directions (e.g., using one or more unit vectors along the one or more directions). In some aspects, when the feedback includes multiple vectors, the UE 120 may project the multiple vectors along the same direction. Alternatively, the UE 120 may project at least some of the multiple vectors along different directions.

[0093] In some respects, these probabilities can be further based, at least in part, on the distribution of the feedback. For example, when the feedback comprises multiple scalars, the UE 120 can adjust these probabilities, at least in part, based on the distribution of these scalars. Accordingly, in one example, when the cumulative distribution function (CDF) associated with multiple feedbacks is 0.0 at -1, the UE 120 can quantize one of the multiple feedbacks with a value of 0.6 to 1 with a probability of 0.95 (or 95%), and quantize it to -1 with a probability of 0.05 (or 5%); however, when the CDF associated with multiple feedbacks is 0.5 at -1, the UE 120 can quantize one of the multiple feedbacks with a value of 0.6 to 1 with a probability of 0.70 (or 70%), and quantize it to -1 with a probability of 0.30 (or 30%). In another example, when the feedback comprises multiple vectors, the UE 120 can adjust these probabilities, at least in part, based on the distribution of the components of these vectors. For example, the probability used to quantize the first component can depend at least in part on the distribution of the first components of these vectors. Similarly, the probability used to quantize the second component can depend at least in part on the distribution of the second components of these vectors. Although the description in this example focuses on vectors with two components, the description is similarly applicable to vectors with additional components, such as three components, four components, etc.

[0094] Additionally or alternatively, these probabilities may be further based, at least in part, on conditions associated with the channel between UE 120 and base station 110. In some aspects, the conditions associated with the channel may include RSRP, CQI, signal-to-noise ratio (SNR), and / or other indicators of channel quality. In one example, UE 120 may use more zero probabilities when the conditions associated with the channel do not meet a threshold. Accordingly, in one example, when the conditions associated with the channel meet a threshold, UE 120 may quantize a feedback value of 0.6 to 1 with a probability of 0.70 (or 70%), to -1 with a probability of 0.10 (or 10%), to -3 with a probability of 0.05 (or 5%), and to 3 with a probability of 0.15 (or 15%).

[0095] However, when the conditions associated with the channel do not meet the threshold, the UE 120 may quantize a feedback value of 0.6 to 1 with a probability of 0.95 (or 95%), to -1 with a probability of 0.50 (or 5%), to -3 with a probability of 0.0 (or 0%), and to 3 with a probability of 0.0 (or 0%). Alternatively, the UE 120 may use additional quantized numbers when the conditions associated with the channel meet the threshold, and use fewer quantized numbers when the conditions associated with the channel do not meet the threshold.

[0096] In some respects, base station 110 can transmit and UE 120 can receive indications of at least one relationship between these probabilities and these distances. For example, the configuration described in conjunction with reference numeral 405 can indicate the at least one relationship. Additionally or alternatively, base station 110 can transmit and UE 120 can receive separate messages (e.g., via RRC signaling) indicating the at least one relationship.

[0097] In some aspects, the at least one relationship may include a formula and / or other algorithm that takes these distances as input and provides these probabilities as output. In some aspects, the at least one relationship may further accept conditions associated with the channel between UE 120 and base station 110 and / or the distribution of the feedback as input.

[0098] In some respects, at least one relationship between these probabilities and distances may be pre-configured. For example, this at least one relationship may be defined in a 3GPP specification and / or another standard. Accordingly, in some respects, UE 120 may be programmed (and / or otherwise pre-configured) with this at least one relationship. Additionally or alternatively, base station 110 may transmit indications of at least one relationship from a plurality of pre-configured relationships for use by UE 120. For example, base station 110 may transmit one or more indexes indicating at least one relationship from a relationship table in a 3GPP specification and / or another standard for use by UE 120.

[0099] Additionally or alternatively, base station 110 may transmit and UE 120 may receive an indication of a quantized digit. For example, the configuration described above in conjunction with reference to reference numeral 405 may indicate a quantized digit. Additionally or alternatively, base station 110 may transmit and UE 120 may receive a separate message indicating a quantized digit (e.g., via RRC signaling). In some aspects, the quantized digit may be fixed. Alternatively, the quantized digit may be dynamic. For example, base station 110 may provide a formula and / or other algorithm that accepts these feedbacks as inputs and provides a quantized digit as an output. In some aspects, the formula and / or other algorithm may further accept conditions associated with the channel between UE 120 and base station 110 and / or the distribution of the feedback as inputs.

[0100] In some respects, the quantized number may be pre-configured. For example, the quantized number may be defined in a 3GPP specification and / or another standard. Accordingly, in some respects, the UE 120 may be programmed (and / or otherwise pre-configured) with these quantized numbers. Additionally or alternatively, the base station 110 may transmit indications of these quantized numbers from a plurality of pre-configured quantized numbers for use by the UE 120. For example, the base station 110 may transmit one or more indices indicating a set of quantized numbers in a table of sets of quantized numbers from a 3GPP specification and / or another standard for use by the UE 120.

[0101] As shown in conjunction with reference numeral 415, UE 120 can transmit and base station 110 can receive quantized values. Base station 110 can determine an update at least in part based on the quantized values. In some aspects, base station 110 can determine the update at least in part based on aggregating quantized values ​​from multiple UEs (e.g., as in conjunction with...). Figure 3 (As described).

[0102] Therefore, base station 110 can update the machine learning component at least in part based on this update. In some aspects, base station 110 can determine multiple updates (e.g., at least in part based on aggregating quantized values ​​from multiple groups of UEs, where each group includes one or more UEs) and aggregate the multiple updates to determine a global update for the machine learning component. In some aspects, Example 400 can be recursive, wherein base station 110 retransmits the updated machine learning component for additional training by UE 120 and / or other UEs in joint learning.

[0103] By using, such as combination Figure 4 The described technique involves UE 120 quantizing the feedback using randomization with probabilities at least partially based on the corresponding distances between one or more values ​​of the feedback and multiple quantized numbers. Accordingly, base station 110 more accurately aggregates feedback from multiple UEs, including UE 120. As a result, UE 120 and base station 110 suffer lower network and memory overhead by using quantization during joint learning, without incurring significant accuracy loss during joint learning.

[0104] As indicated above, Figure 4 This is provided merely as an example. Other examples may differ from those provided. Figure 4 The example described.

[0105] Figure 5 This is a diagram illustrating an example process 500 performed, for example, by a client device according to this disclosure. Example process 500 is wherein a client device (e.g., UE 120, ...) Figure 7 An example of an equipment 700 and / or another client device, such as a tablet, laptop, or desktop computer, performing operations associated with transmission via quantized feedback in joint learning with randomization.

[0106] like Figure 5 As shown, in some aspects, process 500 may include determining feedback associated with the applied machine learning component, at least in part, based on the application of the machine learning component (box 510). For example, the client device (e.g., using...) Figure 7 The determining component 708 described herein can determine the feedback associated with the machine learning component, at least in part, based on the applied machine learning component, as described herein.

[0107] like Figure 5 As further shown, in some aspects, process 500 may include transmitting quantized values ​​at least in part based on this feedback (box 520). For example, a client device (e.g., using transmission component 704, such as...) Figure 7The quantized values ​​(as described herein) can be transmitted at least in part based on this feedback, as described herein. In some aspects, the quantized values ​​are determined using randomization with probabilities of corresponding distances between one or more values, at least in part based on this feedback, and a plurality of quantized numbers.

[0108] Process 500 may include additional aspects, such as any single aspect or any combination of aspects described below and / or in conjunction with one or more other processes described elsewhere herein.

[0109] In the first aspect, the machine learning component includes at least one neural network.

[0110] In the second aspect, either alone or in combination with the first aspect, the feedback includes at least one weight.

[0111] In the third aspect, the feedback includes at least one vector, either alone or in combination with one or more of the first and second aspects.

[0112] In the fourth aspect, either alone or in combination with one or more of the first to third aspects, the quantized value is at least partially based on a component of the at least one vector.

[0113] In the fifth aspect, either alone or in combination with one or more of the first to fourth aspects, the quantized value is at least partially based on two or more components of the at least one vector.

[0114] In the sixth aspect, either alone or in combination with one or more of the first to fifth aspects, the quantized value is at least partially based on the projection of the at least one vector.

[0115] In the seventh aspect, either alone or in combination with one or more of the first to sixth aspects, the probability is further based at least in part on the distribution of the feedback.

[0116] In the eighth aspect, either alone or in combination with one or more of the first to seventh aspects, the probability is further based, at least in part, on conditions associated with the channel between the client device and the server device.

[0117] In the ninth aspect, alone or in combination with one or more of the first to eighth aspects, process 500 further includes receiving (e.g., using...) Figure 7 The depicted receiving component 702) indicates at least one relationship between probability and distance.

[0118] In the tenth aspect, either alone or in combination with one or more of the first to ninth aspects, at least one relationship between probability and distance is pre-configured.

[0119] although Figure 5An example box of process 500 is shown, but in some respects, process 500 may include... Figure 5 The boxes depicted in the process are compared to additional boxes, fewer boxes, different boxes, or boxes arranged differently. Additionally or alternatively, two or more boxes in process 500 can be executed in parallel.

[0120] Figure 6 This is a diagram illustrating an example process 600 performed by, for example, a server device according to this disclosure. Example process 600 is a server device (e.g., base station 110, ...). Figure 8 An example of an equipment 800 and / or another server device, such as one or more server computers in a server cluster and / or at least a portion of the core network supporting base station 110, performing operations associated with receiving quantized feedback in joint learning with randomization.

[0121] like Figure 6 As shown, in some aspects, process 600 may include sending a command to a client device (e.g., UE 120, ...). Figure 7 The device 700, and / or another client device, such as a tablet, laptop, or desktop computer, transmits the configuration associated with the machine learning component (box 610). For example, a server device (e.g., using...) Figure 8 The described transmission component 804 can transmit configuration associated with the machine learning component to a client device, as described herein. In some aspects, the machine learning component accepts one or more inputs to generate one or more outputs.

[0122] like Figure 6 As further shown, in some aspects, process 600 may include receiving quantized values ​​(box 620) based at least in part on feedback from a client device to which machine learning components have been applied. For example, a server device (e.g., using receiving component 802, such as...) Figure 8 The quantized values ​​(described herein) can receive quantized values ​​based at least in part on feedback from client devices to which machine learning components have been applied. In some aspects, the quantized values ​​are based at least in part on randomization having probabilities of corresponding distances between one or more values ​​of feedback and a plurality of quantized numbers.

[0123] Process 600 may include additional aspects, such as any single aspect or any combination of aspects described below and / or in conjunction with one or more other processes described elsewhere herein.

[0124] In the first aspect, the machine learning component includes at least one neural network.

[0125] In the second aspect, either alone or in combination with the first aspect, the feedback includes at least one weight.

[0126] In the third aspect, the feedback includes at least one vector, either alone or in combination with one or more of the first and second aspects.

[0127] In the fourth aspect, either alone or in combination with one or more of the first to third aspects, the quantized value is at least partially based on a component of the at least one vector.

[0128] In the fifth aspect, either alone or in combination with one or more of the first to fourth aspects, the quantized value is at least partially based on two or more components of the at least one vector.

[0129] In the sixth aspect, either alone or in combination with one or more of the first to fifth aspects, the quantized value is at least partially based on the projection of the at least one vector.

[0130] In the seventh aspect, either alone or in combination with one or more of the first to sixth aspects, the probability is further based at least in part on the distribution of the feedback.

[0131] In the eighth aspect, either alone or in combination with one or more of the first to seventh aspects, the probability is further based, at least in part, on conditions associated with the channel between the client device and the server device.

[0132] In the ninth aspect, alone or in combination with one or more of the first to eighth aspects, process 600 further includes transmitting (e.g., using transmission component 804) an indication of at least one relationship between probability and distance to a client device.

[0133] In the tenth aspect, either alone or in combination with one or more of the first to ninth aspects, at least one relationship between probability and distance is pre-configured.

[0134] although Figure 6 An example box of process 600 is shown, but in some respects, process 600 may include... Figure 6 The boxes depicted in the process are compared to additional boxes, fewer boxes, different boxes, or boxes arranged differently. Additionally or alternatively, two or more boxes in process 600 can be executed in parallel.

[0135] Figure 7This is a block diagram of an example device 700 for wireless communication. Device 700 may be a client device, or a client device may include device 700. In some aspects, device 700 includes a receiving component 702 and a transmitting component 704, which may be in communication with each other (e.g., via one or more buses and / or one or more other components). As shown, device 700 may use the receiving component 702 and the transmitting component 704 to communicate with another device 706 (such as a server device, UE, base station, or another wireless communication device). As further shown, device 700 may include a determining component 708 and other examples.

[0136] In some respects, device 700 can be configured to perform the functions described in this article. Figure 4 One or more operations described herein. Additionally or alternatively, equipment 700 may be configured to perform one or more processes described herein (such as...). Figure 5 Process 500) or a combination thereof. In some aspects, equipment 700 and / or Figure 7 One or more components shown may include the above combination Figure 2 One or more components of the described UE. Additionally or alternatively, Figure 7 One or more components shown can be combined as described above. Figure 2 Implemented within one or more of the described components. Additionally or alternatively, one or more components in the set of components may be implemented at least partially as software stored in memory. For example, a component (or a portion of a component) may be implemented as instructions or code stored in a non-transitory computer-readable medium and may be executed by a controller or processor to perform the function or operation of that component.

[0137] Receiver 702 may receive communications (such as reference signals, control information, data communications, or combinations thereof) from device 706. Receiver 702 may provide the received communications to one or more other components of device 700. In some aspects, receiver 702 may perform signal processing on the received communications (such as filtering, amplification, demodulation, analog-to-digital conversion, demultiplexing, deinterleaving, demapping, equalization, interference cancellation, or decoding, and other examples), and may provide the processed signal to one or more other components of device 700. In some aspects, receiver 702 may include combinations of the above. Figure 2 The described UE includes one or more antennas, demodulators, MIMO detectors, receiver processors, controllers / processors, memory, or combinations thereof.

[0138] The transmission component 704 can transmit communications (such as reference signals, control information, data communications, or combinations thereof) to the device 706. In some aspects, one or more other components of the device 700 can generate communications and provide the generated communications to the transmission component 704 for transmission to the device 706. In some aspects, the transmission component 704 can perform signal processing (such as filtering, amplification, modulation, digital-to-analog conversion, multiplexing, interleaving, mapping, encoding, etc.) on the generated communications and can transmit the processed signals to the device 706. In some aspects, the transmission component 704 can include combinations of the above. Figure 2 The described UE includes one or more antennas, modulators, transmit MIMO processors, transmit processors, controllers / processors, memories, or combinations thereof. In some aspects, the transmit component 704 may coexist with the receive component 702 in a transceiver.

[0139] In some aspects, determining component 708 can be at least partially based on the application of a machine learning component to determine the feedback associated with that machine learning component. In some aspects, determining component 708 may include a combination of the above. Figure 2 The described UE includes a receiving processor, a transmitting processor, a controller / processor, a memory, or a combination thereof. Additionally, the transmission component 704 may transmit (e.g., to a server device, such as device 706) the quantized value based at least in part on this feedback. In some aspects, the determining component 708 may determine the quantized value using randomization having probabilities of corresponding distances between one or more values ​​based at least in part on this feedback and a plurality of quantized numbers.

[0140] In some aspects, receiving component 702 may (e.g., from device 706) receive an indication of at least one relationship between probability and distance. Additionally or alternatively, this at least one relationship may be pre-configured for equipment 700.

[0141] Figure 7 The number and arrangement of components shown are provided as an example. In practice, different arrangements may exist. Figure 7 The components shown are compared to additional components, fewer components, different components, or components arranged differently. Furthermore, Figure 7 The two or more components shown can be implemented within a single component, or Figure 7 The single component shown can be implemented as multiple distributed components. Additionally or alternatively, Figure 7 The collection of components shown (e.g., one or more components) can be executed as described by Figure 7 The other set of components shown in the diagram performs one or more functions.

[0142] Figure 8This is a block diagram of an example device 800 for wireless communication. Device 800 may be a server device, or a server device may include device 800. In some aspects, device 800 includes a receiving component 802 and a transmitting component 804, which may be in communication with each other (e.g., via one or more buses and / or one or more other components). As shown, device 800 may use the receiving component 802 and the transmitting component 804 to communicate with another device 806 (such as a client device, UE, base station, or another wireless communication device). As further shown, device 800 may include a quantization component 808 and other examples.

[0143] In some respects, Equipment 800 can be configured to perform the actions described in this article. Figure 4 One or more operations described herein. Additionally or alternatively, equipment 800 may be configured to perform one or more processes described herein (such as...). Figure 6 Process 600) or a combination thereof. In some aspects, equipment 800 and / or Figure 8 One or more components shown may include the above combination Figure 2 One or more components of the described base station. Additional or alternative. Figure 8 One or more components shown can be combined as described above. Figure 2 Implemented within one or more of the described components. Additionally or alternatively, one or more components in the set of components may be implemented at least partially as software stored in memory. For example, a component (or a portion of a component) may be implemented as instructions or code stored in a non-transitory computer-readable medium and may be executed by a controller or processor to perform the function or operation of that component.

[0144] Receiver 802 may receive communications (such as reference signals, control information, data communications, or combinations thereof) from device 806. Receiver 802 may provide the received communications to one or more other components of device 800. In some aspects, receiver 802 may perform signal processing on the received communications (such as filtering, amplification, demodulation, analog-to-digital conversion, demultiplexing, deinterleaving, demapping, equalization, interference cancellation, or decoding, and other examples), and may provide the processed signal to one or more other components of device 800. In some aspects, receiver 802 may include combinations of the above. Figure 2 The described base station includes one or more antennas, demodulators, MIMO detectors, receiver processors, controllers / processors, memory, or combinations thereof.

[0145] The transmission component 804 can transmit communications (such as reference signals, control information, data communications, or combinations thereof) to the device 806. In some aspects, one or more other components of the device 800 can generate communications and provide the generated communications to the transmission component 804 for transmission to the device 806. In some aspects, the transmission component 804 can perform signal processing (such as filtering, amplification, modulation, digital-to-analog conversion, multiplexing, interleaving, mapping, encoding, etc.) on the generated communications and can transmit the processed signals to the device 806. In some aspects, the transmission component 804 can include combinations of the above. Figure 2 The described base station includes one or more antennas, modulators, transmit MIMO processors, transmit processors, controllers / processors, memory, or combinations thereof. In some aspects, the transmit component 804 may coexist with the receive component 802 in a transceiver.

[0146] The transmission component 804 can transmit (e.g., to a client device, such as device 806) a configuration associated with the machine learning component, wherein the machine learning component accepts one or more inputs to generate one or more outputs. Accordingly, the receiving component 802 can receive quantized values ​​based at least in part on feedback from a device to which the machine learning component has been applied, wherein the quantized values ​​are based at least in part on randomization having probabilities based at least in part on the distances between one or more values ​​of the feedback and a plurality of quantized numbers.

[0147] In some aspects, transmission component 804 may transmit to device 806 an indication of at least one relationship between probability and distance. For example, quantization component 808 may determine at least one relationship (e.g., based at least in part on the expected distribution of conditions and / or feedback associated with the channel between the client device and the server device). In some aspects, quantization component 808 may include the above combination. Figure 2 The described base station includes a receiving processor, a transmitting processor, a controller / processor, a memory, or a combination thereof. Additionally or alternatively, this at least one relationship may be pre-configured.

[0148] Figure 8 The number and arrangement of components shown are provided as an example. In practice, different arrangements may exist. Figure 8 The components shown are compared to additional components, fewer components, different components, or components arranged differently. Furthermore, Figure 8 The two or more components shown can be implemented within a single component, or Figure 8 The single component shown can be implemented as multiple distributed components. Additionally or alternatively, Figure 8 The collection of components shown (e.g., one or more components) can be executed as described by Figure 8 The other set of components shown in the diagram performs one or more functions.

[0149] The following provides an overview of some aspects of this disclosure:

[0150] Aspect 1: A method for performing wireless communication by a client device, comprising: determining feedback associated with a machine learning component based at least in part on an applied machine learning component; and transmitting a quantized value based at least in part on the feedback, wherein the quantized value is determined using randomization having probabilities of corresponding distances between one or more values ​​based at least in part on the feedback and a plurality of quantized numbers.

[0151] Aspect 2: The method of aspect 1, wherein the machine learning component includes at least one neural network.

[0152] Aspect 3: The method of any one of Aspects 1 to 2, wherein the feedback includes at least one weight.

[0153] Aspect 4: The method of any one of Aspects 1 to 3, wherein the feedback includes at least one vector.

[0154] Aspect 5: The method of aspect 4, wherein the quantized value is at least partially based on a component of the at least one vector.

[0155] Aspect 6: The method of aspect 4, wherein the quantized value is based at least in part on two or more components of the at least one vector.

[0156] Aspect 7: The method of any one of Aspects 4 to 6, wherein the quantized value is at least partially based on the projection of the at least one vector.

[0157] Aspect 8: The method of any of Aspects 1 to 7, wherein the probability is further based at least in part on the distribution of the feedback.

[0158] Aspect 9: The method of any of Aspects 1 to 8, wherein the probability is further based at least in part on conditions associated with the channel between the client device and the server device.

[0159] Aspect 10: The method of any of aspects 1 to 9 further includes: receiving an indication of at least one relationship between probability and distance.

[0160] Aspect 11: The method of any one of Aspects 1 to 10, wherein at least one relationship between probability and distance is pre-configured.

[0161] Aspect 12: A method of performing wireless communication by a server device, comprising: transmitting to a client device a configuration associated with a machine learning component, wherein the machine learning component accepts one or more inputs to generate one or more outputs; and receiving quantized values ​​based at least in part on feedback from a client device to which the machine learning component has been applied, wherein the quantized values ​​are based at least in part on randomization having probabilities of corresponding distances between one or more values ​​of the feedback and a plurality of quantized numbers.

[0162] Aspect 13: The method of aspect 12, wherein the machine learning component includes at least one neural network.

[0163] Aspect 14: The method of any of Aspects 12 to 13, wherein the feedback includes at least one weight.

[0164] Aspect 15: The method of any of Aspects 12 to 14, wherein the feedback includes at least one vector.

[0165] Aspect 16: The method of aspect 15, wherein the quantized value is at least partially based on a component of the at least one vector.

[0166] Aspect 17: The method of aspect 15, wherein the quantized value is based at least in part on two or more components of the at least one vector.

[0167] Aspect 18: The method of any one of aspects 15 to 17, wherein the quantized value is at least partially based on the projection of the at least one vector.

[0168] Aspect 19: The method of any of Aspects 12 to 18, wherein the probability is further based at least in part on the distribution of the feedback.

[0169] Aspect 20: The method of any one of Aspects 12 to 19, wherein the probability is further based at least in part on conditions associated with the channel between the client device and the server device.

[0170] Aspect 21: The method of any of Aspects 12 to 20 further includes: transmitting to a client device an indication of at least one relationship between probability and distance.

[0171] Aspect 22: The method of any one of Aspects 12 to 21, wherein at least one relationship between probability and distance is pre-configured.

[0172] Aspect 23: An apparatus for wireless communication at a device, comprising: a processor; a memory coupled to the processor; and instructions stored in the memory and executable by the processor to cause the apparatus to perform methods as described in one or more of aspects 1-11.

[0173] Aspect 24: An apparatus for wireless communication, comprising a memory and one or more processors coupled to the memory, the one or more processors being configured to perform methods as described in one or more aspects of aspects 1-11.

[0174] Aspect 25: An apparatus for wireless communication, comprising at least one means for performing a method as described in one or more aspects of aspects 1-11.

[0175] Aspect 26: A non-transient computer-readable medium storing code for wireless communication, the code including instructions executable by a processor to perform methods as described in one or more aspects of aspects 1-11.

[0176] Aspect 27: A non-transient computer-readable medium storing a set of instructions for wireless communication, the set of instructions including one or more instructions which, when executed by one or more processors of a device, cause the device to perform methods as described in one or more aspects of aspects 1-11.

[0177] Aspect 28: An apparatus for wireless communication at a device, comprising: a processor; a memory coupled to the processor; and instructions stored in the memory and executable by the processor to cause the apparatus to perform methods as described in one or more aspects of aspects 12-22.

[0178] Aspect 29: An apparatus for wireless communication, comprising a memory and one or more processors coupled to the memory, the one or more processors being configured to perform methods as described in one or more aspects of aspects 12-22.

[0179] Aspect 30: An apparatus for wireless communication, comprising at least one means for performing a method as described in one or more aspects of aspects 12-22.

[0180] Aspect 31: A non-transient computer-readable medium storing code for wireless communication, the code including instructions executable by a processor to perform methods as described in one or more aspects of aspects 12-22.

[0181] Aspect 32: A non-transient computer-readable medium storing a set of instructions for wireless communication, the set of instructions including one or more instructions which, when executed by one or more processors of a device, cause the device to perform methods as described in one or more aspects of aspects 12-22.

[0182] The foregoing disclosure provides explanations and descriptions, but is not intended to be exhaustive or to limit the aspects to the precise forms disclosed. Modifications and variations may be made in light of the foregoing disclosure or may be obtained through practice.

[0183] As used herein, the term "component" is intended to be broadly interpreted as hardware and / or a combination of hardware and software. "Software" should be broadly interpreted as instructions, instruction sets, code, code segments, program code, programs, subroutines, software modules, applications, software applications, software packages, routines, subroutines, objects, executables, threads of execution, procedures, and / or functions, whether referred to as software, firmware, middleware, microcode, hardware description languages, or other terms. As used herein, a "processor" is implemented in hardware and / or a combination of hardware and software. It will be apparent that the systems and / or methods described herein can be implemented in various forms of hardware and / or combinations of hardware and software. The actual dedicated control hardware or software code used to implement these systems and / or methods is not limited in any way. Therefore, the operation and behavior of these systems and / or methods are described herein without reference to any specific software code, as those skilled in the art will understand that the software and hardware can be designed to implement these systems and / or methods, at least in part, based on the description herein.

[0184] As used in this article, depending on the context, "meeting the threshold" can mean a value greater than the threshold, greater than or equal to the threshold, less than the threshold, less than or equal to the threshold, equal to the threshold, not equal to the threshold, etc.

[0185] Although specific combinations of features are described in the claims and / or disclosed in the specification, these combinations are not intended to limit the disclosure of aspects. Many of these features may be combined in ways not specifically described in the claims and / or disclosed in the specification. The disclosure of aspects includes each dependent claim in combination with each other claim in the claim set. As used herein, the phrase “at least one of” refers to any combination of these items, including a single member. As an example, “at least one of a, b, or c” is intended to cover: a, b, c, a+b, a+c, b+c, and a+b+c, as well as any combination having multiple identical elements (e.g., a+a, a+a+a, a+a+b, a+a+c, a+b+b, a+c+c, b+b, b+b+b, b+b+c, c+c, and c+c+c, or any other ordering of a, b, and c).

[0186] The elements, actions, or instructions used herein should not be construed as critical or necessary unless explicitly stated otherwise. Furthermore, as used herein, the articles “a” and “a certain” are intended to include one or more items and may be used interchangeably with “one or more.” Additionally, as used herein, the article “the” is intended to include one or more items referenced in conjunction with the article “the” and may be used interchangeably with “one or more.” Furthermore, as used herein, the terms “set” and “group” are intended to include one or more items and may be used interchangeably with “one or more.” In cases where only one item is intended, the phrase “only one” or similar language is used. Moreover, as used herein, the terms “have,” “contain,” “include,” etc., are intended to be open-ended terms that do not limit the elements they modify (e.g., the element “has” A may also have B). Furthermore, the phrase “based on” is intended to mean “at least partially based on” unless otherwise explicitly stated. Moreover, as used herein, the term “or” is intended to be inclusive when used in a sequence and may be used interchangeably with “and / or” unless otherwise explicitly stated (e.g., in combination with “either of” or “only one of”).

Claims

1. An apparatus for wireless communication at a user equipment location, comprising: One or more memory units; as well as One or more processors coupled to the one or more memories, the one or more processors being configured to: Receives configuration associated with machine learning components; The feedback associated with the machine learning component is determined at least in part based on the application of the machine learning component, wherein the feedback includes a plurality of quantized scalar values ​​corresponding to the local gradients or updated weights of the machine learning component; The quantized values ​​are determined at least in part based on the feedback and using probabilistic randomization, wherein the probability is at least in part based on: the corresponding distance between each of the plurality of quantized scalar values ​​of the feedback and each of the plurality of quantized numbers, the distribution of the feedback across the plurality of quantized scalar values, and the conditions associated with the channel between the user equipment and the network node. as well as The quantized value is transmitted to the network node.

2. The apparatus of claim 1, wherein the machine learning component comprises at least one neural network.

3. The apparatus of claim 1, wherein the plurality of quantized scalar values ​​correspond to the updated weights of the machine learning components.

4. The apparatus of claim 1, wherein the plurality of quantized scalar values ​​correspond to at least one vector.

5. The apparatus of claim 4, wherein the quantized value is at least partially based on a component of the at least one vector.

6. The apparatus of claim 4, wherein the quantized value is based at least in part on two or more components of the at least one vector.

7. The apparatus of claim 4, wherein the quantized value is at least partially based on the projection of the at least one vector.

8. The apparatus of claim 1, wherein, The one or more processors are further configured to: Receive an indication of at least one relationship between the probability and the corresponding distance.

9. The apparatus of claim 1, wherein at least one relationship between the probability and the corresponding distance is pre-configured.

10. An apparatus for wireless communication at a network node, comprising: One or more memory units; as well as One or more processors coupled to the one or more memories, the one or more processors being configured to: Transmitting configuration associated with a machine learning component to a user equipment, wherein the machine learning component accepts one or more inputs to generate one or more outputs; and Receive from the user equipment quantized values ​​at least in part based on feedback from the user equipment to which the machine learning component has been applied, wherein the quantized values ​​are at least in part based on probabilistic randomization, wherein the probabilistic randomization is at least in part based on: The corresponding distance between each of the plurality of quantized scalar values ​​in the feedback and each of the plurality of quantized numbers. The feedback is distributed across the plurality of quantized numbers, and Conditions associated with the channel between the user equipment and the network node.

11. The apparatus of claim 10, wherein the machine learning component comprises at least one neural network.

12. The apparatus of claim 10, wherein the plurality of quantized scalar values ​​correspond to the updated weights of the machine learning components.

13. The apparatus of claim 10, wherein the plurality of quantized scalar values ​​correspond to at least one vector.

14. The apparatus of claim 13, wherein the quantized value is at least partially based on a component of the at least one vector.

15. The apparatus of claim 13, wherein the quantized value is based at least in part on two or more components of the at least one vector.

16. The apparatus of claim 13, wherein the quantized value is at least partially based on the projection of the at least one vector.

17. The apparatus of claim 10, wherein the one or more processors are further configured to: The user equipment is transmitted with an indication of at least one relationship between the probability and the corresponding distance.

18. The apparatus of claim 10, wherein at least one relationship between the probability and the corresponding distance is pre-configured.

19. A method for performing wireless communication by a user equipment, comprising: Receives configuration associated with machine learning components; The feedback associated with the machine learning component is determined at least in part based on the application of the machine learning component, wherein the feedback includes a plurality of quantized scalar values ​​corresponding to the local gradients or updated weights of the machine learning component; The quantized value is determined at least in part based on the feedback, wherein the quantized value is determined using probabilistic randomization, wherein the probability is at least in part based on: The corresponding distance between each of the plurality of quantized scalar values ​​and each of the plurality of quantized numbers in the feedback. The feedback is distributed across the plurality of quantized scalar values, and Conditions associated with the channel between the user equipment and the network node; as well as The quantized value is transmitted to the network node.

20. The method of claim 19, wherein the plurality of quantized scalar values ​​correspond to the updated weights of the machine learning components.

21. The method of claim 19, wherein the plurality of quantized scalar values ​​correspond to at least one vector.

22. The method of claim 19, further comprising: Receive an indication of at least one relationship between the probability and the corresponding distance.

23. The method of claim 19, wherein at least one relationship between the probability and the corresponding distance is pre-configured.

24. A method for performing wireless communication by a network node, comprising: The configuration associated with a machine learning component is transmitted to the user equipment, wherein the machine learning component accepts one or more inputs to generate one or more outputs; as well as Receive from the user equipment quantized values ​​at least in part based on feedback from the user equipment to which the machine learning component has been applied, wherein the quantized values ​​are at least in part based on probabilistic randomization, wherein the probabilistic randomization is at least in part based on: The corresponding distance between each of the plurality of quantized scalar values ​​in the feedback and each of the plurality of quantized numbers. The feedback is distributed across the plurality of quantized numbers, and The conditions associated with the channel between the user equipment and the network node, and The plurality of quantized numbers includes a multi-bit representation corresponding to the plurality of numbers.

25. The method of claim 24, wherein the machine learning component comprises at least one neural network.

26. The method of claim 24, wherein the plurality of quantized scalar values ​​correspond to the updated weights of the machine learning components.

27. The method of claim 24, wherein the plurality of quantized scalar values ​​correspond to at least one vector.

28. The method of claim 27, wherein the quantified value is based at least in part on at least one of the following: One component of the at least one vector, or The projection of the at least one vector.

29. The method of claim 24, further comprising: The user equipment is transmitted with an indication of at least one relationship between the probability and the corresponding distance.

30. The method of claim 24, wherein at least one relationship between the probability and the corresponding distance is pre-configured.

Images