Determining parameters of multiple models of wireless communication system

Abstract

Apparatuses, methods, and systems for determining parameters of multiple models of a wireless communication system are disclosed. A method (600) includes determining (602), at a first device, a set of parameters using a first set of information, the set of parameters including first information corresponding to a first model and a second model. The method (600) comprises sending (604) a second set of information to a second device, the second set of information comprising second information for the first model or the second model.

Description

Technical Field

[0001] The subject matter disclosed herein relates generally to wireless communications, and more particularly to determining parameters of multiple models of a wireless communication system. Background Art

[0002] In some wireless communication systems, models may be used in the wireless communication system. The transmission of data used to train such models may use a large amount of resources. Summary of the invention

[0003] Methods for determining parameters of multiple models are disclosed. Apparatus and systems also perform the functions of these methods. One embodiment of a method includes: determining a parameter set using a first information set at a first device, the parameter set including first information corresponding to a first model and a second model. In some embodiments, the method includes: sending a second information set to a second device, the second information set including second information for the first model or the second model.

[0004] An apparatus for determining parameters of a plurality of models includes a processor. In some embodiments, the apparatus includes a memory coupled to the processor, the processor being configured to cause the apparatus to: determine a parameter set using a first information set, the parameter set including first information corresponding to a first model and a second model; and send a second information set to a second device, the second information set including second information for the first model or the second model.

[0005] Another embodiment of a method for determining parameters of multiple models includes: receiving, at a second device, a set of information from a first device, the set of information including first information corresponding to a first model and a second model. In some embodiments, the method includes: determining a third model using the first information. In some embodiments, the method includes: generating an output based on the third model and the first data set.

[0006] Another apparatus for determining parameters of a plurality of models includes a processor. In some embodiments, the apparatus includes a memory coupled to the processor, the processor being configured to cause the apparatus to: receive an information set from a first device, the information set including first information corresponding to a first model and a second model; determine a third model using the first information; and generate an output based on the third model and the first data set. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] The above embodiments will be described in more detail with reference to the specific embodiments shown in the accompanying drawings. It should be understood that these drawings depict only some embodiments and are therefore not to be considered limiting of scope, and the embodiments will be described and explained with additional specificity and detail using the accompanying drawings, in which:

[0008] Figure 1is a schematic block diagram illustrating one embodiment of a wireless communication system for determining parameters of a plurality of models;

[0009] Figure 2 is a schematic block diagram illustrating one embodiment of an apparatus that may be used to determine parameters of a plurality of models;

[0010] Figure 3 is a schematic block diagram illustrating one embodiment of an apparatus that may be used to determine parameters of a plurality of models;

[0011] Figure 4 is a schematic block diagram illustrating one embodiment of a wireless network;

[0012] Figure 5 is a schematic block diagram illustrating one embodiment of a system using a two-sided model;

[0013] Figure 6 is a flow chart illustrating one embodiment of a method for determining parameters of a plurality of models; and

[0014] Figure 7 is a flow chart illustrating another embodiment of a method for determining parameters of a plurality of models. DETAILED DESCRIPTION

[0015] As will be appreciated by those skilled in the art, aspects of the embodiments may be embodied as systems, devices, methods or program products. Therefore, the embodiments may take the form of a complete hardware embodiment, a complete software embodiment (including firmware, resident software, microcode, etc.), or an embodiment combining software and hardware aspects, which are generally referred to herein as "circuit", "module" or "system". In addition, the embodiments may take the form of a program product, which is embodied in one or more computer-readable storage devices, which store machine-readable code, computer-readable code and / or program code, hereinafter referred to as code. The storage device may be tangible, non-transient and / or non-transmitting. The storage device may not embody a signal. In a specific embodiment, the storage device only uses a signal to access the code.

[0016] Some of the functional units described in this specification may be labeled as modules to more particularly emphasize their implementation independence. For example, a module may be implemented as a hardware circuit, including a custom very large scale integrated circuit (VLSI) circuit or gate array, an off-the-shelf semiconductor (such as a logic chip, transistor, or other discrete component). A module may also be implemented in a programmable hardware device, such as a field programmable gate array, programmable array logic, a programmable logic device, etc.

[0017] Modules can also be implemented with code and / or software for various types of processors to perform. The code module of an identification can, for example, include one or more physical or logical blocks of executable code, which can, for example, be organized as objects, processes or functions. However, the executable program of the identified module does not need to be physically located together, but can include different instructions stored in different locations, which, when logically connected together, include the module and realize the described purpose of the module.

[0018] In fact, a code module can be a single instruction or many instructions, and can even be distributed across several different code segments, between different programs, and between several memory devices. Similarly, operational data can be identified and illustrated within a module, and can be embodied in any suitable form and organized in any suitable type of data structure. The operational data can be collected as a single data set, or can be distributed in different locations, including on different computer-readable storage devices. In the case where a module or a portion of a module is implemented in software, the software portion is stored on one or more computer-readable storage devices.

[0019] Any combination of one or more computer readable media may be utilized. The computer readable medium may be a computer readable storage medium. The computer readable storage medium may be a storage device storing code. The storage device may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, holographic, micromechanical, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing.

[0020] More specific examples of storage devices (a non-exhaustive list) include the following: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer-readable storage medium may be any tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device.

[0021] The code for performing the operations of the embodiments can be any number of lines and can be written in any combination of one or more programming languages, including object-oriented programming languages ​​(such as Python, Ruby, Java, Smalltalk, C++, etc.), as well as traditional procedural programming languages ​​(such as "C" programming language, etc.) and / or machine languages ​​(such as assembly language). The code can be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter case, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or a connection to an external computer (e.g., via the Internet using an Internet service provider) can be implemented.

[0022] References throughout this specification to "one embodiment," "an embodiment," or similar language indicate that a particular feature, structure, or characteristic described in conjunction with the embodiment is included in at least one embodiment. Thus, unless expressly provided otherwise, the phrases "in one embodiment," "in an embodiment," and similar language throughout this specification may, but do not necessarily, refer to the same embodiment, but rather to "one or more but not all embodiments." Unless expressly provided otherwise, the terms "including," "comprising," "having," and variations thereof, mean "including but not limited to." Unless expressly provided otherwise, an enumerated list of items does not imply that any or all items are mutually exclusive. Unless expressly provided otherwise, the terms "a," "an," and "the" also mean "one or more."

[0023] In addition, the features, structures or characteristics of the described embodiments may be combined in any suitable manner. In the following description, many specific details are provided, such as examples of programming, software modules, user selections, network transactions, database queries, database structures, hardware modules, hardware circuits, hardware chips, etc., to provide a comprehensive understanding of the embodiments. However, those skilled in the relevant art will recognize that the embodiments may be practiced without one or more specific details or using other methods, components, materials, etc. In other cases, well-known structures, materials or operations are not shown or described in detail to avoid confusing the various aspects of the embodiments.

[0024] The following reference is made to the schematic flow chart and / or schematic block diagram of the method, device, system and program product according to the embodiment to describe various aspects of the embodiment. It should be understood that each block of the schematic flow chart and / or schematic block diagram and the combination of blocks in the schematic flow chart and / or schematic block diagram can be realized by code. The code can be provided to a processor of a general-purpose computer, a special-purpose computer or other programmable data processing device to produce a machine, so that instructions executed via the processor of the computer or other programmable data processing devices can create a component for realizing the function / action specified in one or more schematic flow chart blocks and / or schematic block diagram blocks.

[0025] The code may also be stored in a storage device that can direct a computer, other programmable data processing apparatus, or other device to operate in a specific manner so that the instructions stored in the storage device produce an article of manufacture including instructions for implementing the functions / actions specified in one or more schematic flowchart blocks and / or schematic block diagram blocks.

[0026] The code may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be executed on the computer, other programmable apparatus, or other device to produce a computer-implemented process, such that the code executed on the computer or other programmable apparatus provides a process for implementing the functions / actions specified in one or more flowchart blocks and / or block diagram blocks.

[0027] The schematic flow charts and / or schematic block diagrams in the figures illustrate the possible implementation architectures, functions and operations of the devices, systems, methods and program products according to various embodiments. In this regard, each block in the schematic flow charts and / or schematic block diagrams may represent a module, segment or portion of a code, which includes one or more executable instructions for implementing the code of (multiple) specified logical functions.

[0028] It should also be noted that in some alternative implementations, the functions noted in the blocks may not appear in the order noted in the figures. For example, two blocks shown in succession may in fact be executed substantially concurrently, or the blocks may sometimes be executed in reverse order, depending on the functions involved. Other steps and methods may be envisioned that are equivalent in function, logic, or effect to one or more blocks or portions thereof of the illustrated figures.

[0029] Although various arrow types and line types can be adopted in flow chart and / or block diagram, it should be understood that they do not limit the scope of corresponding embodiment.In fact, some arrows or other connectors can be used to indicate the logical flow of the described embodiment only.For example, the waiting or monitoring period of the unspecified duration between the enumeration steps of the described embodiment can be indicated by an arrow.It will also be noted that each block of block diagram and / or flow chart and the combination of blocks in block diagram and / or flow chart can be realized by a system based on special hardware or a combination of special hardware and code that performs a specified function or action.

[0030] The description of elements in each figure may refer to elements of subsequent figures. Like reference numerals refer to like elements throughout the drawings, including alternative embodiments of the like elements.

[0031] Figure 1 An embodiment of a wireless communication system 100 for determining parameters of a plurality of models is depicted. In one embodiment, the wireless communication system 100 includes a remote unit 102 and a network unit 104. Although Figure 1 A specific number of remote units 102 and network units 104 are depicted in FIG. 1 , but one skilled in the art will recognize that any number of remote units 102 and network units 104 may be included in the wireless communication system 100.

[0032] In one embodiment, the remote unit 102 may include a computing device such as a desktop computer, a laptop computer, a personal digital assistant (PDA), a tablet computer, a smart phone, a smart TV (e.g., a TV connected to the Internet), a set-top box, a game console, a security system (including a security camera), a vehicle computer, a network device (e.g., a router, a switch, a modem), an aircraft, a drone, etc. In some embodiments, the remote unit 102 includes a wearable device such as a smart watch, a fitness band, an optical head-mounted display, etc. In addition, the remote unit 102 may be referred to as a user unit, a mobile station, a mobile station, a user, a terminal, a mobile terminal, a fixed terminal, a user station, a user equipment (UE), a user terminal, a device, or other terms used in the art. The remote unit 102 may communicate directly with one or more of the network units 104 via UL communication signals. In some embodiments, the remote unit 102 may communicate directly with other remote units 102 via side link communications.

[0033] The network elements 104 may be distributed over a geographic area. In some embodiments, the network elements 104 may also be referred to as and / or may include one or more of the following: access point, access terminal, base station, base station, location server, core network (CN), radio network entity, node B, evolved node B (eNB), 5G node B (gNB), home node B, relay node, device, core network, air server, radio access node, access point (AP), new radio (NR), network entity, access and mobility management function (AMF), unified data management (UDM), unified database (UDR), UDM / UDR, policy control function (PCF), radio access network (RAN), network slice selection function (NSSF), operation, maintenance and management (OAM), session management function (SMF), user plane function (UPF), application function, authentication server function (AUSF), security anchor function (SEAF), trusted non-3GPP gateway function (TNGF) or any other term used in the art. The network element 104 is typically part of a radio access network, which includes one or more controllers communicatively coupled to one or more corresponding network elements 104. The radio access network is typically communicatively coupled to one or more core networks, which may be coupled to other networks, such as the Internet and the public switched telephone network, among other networks. These and other elements of the radio access network and core network are not shown, but are generally known to those of ordinary skill in the art.

[0034] In one implementation, the wireless communication system 100 conforms to the NR protocol standardized in the Third Generation Partnership Project (3GPP), wherein the network unit 104 transmits on the downlink (DL) using an orthogonal frequency division multiplexing (OFDM) modulation scheme, and the remote unit 102 transmits on the uplink (UL) using a single carrier frequency division multiple access (SC-FDMA) scheme or an OFDM scheme. However, more generally, the wireless communication system 100 may implement some other open or proprietary communication protocol, such as WiMAX, an Institute of Electrical and Electronics Engineers (IEEE) 802.11 variant, a Global System for Mobile Communications (GSM), a General Packet Radio Service (GPRS), a Universal Mobile Telecommunications System (UMTS), a Long Term Evolution (LTE) variant, a Code Division Multiple Access 2000 (CDMA2000), ZigBee, Sigfox, etc. The present disclosure is not intended to be limited to any particular wireless communication system architecture or protocol implementation.

[0035] The network unit 104 can serve multiple remote units 102 within a service area (eg, a cell or cell sector) via wireless communication links. The network unit 104 sends DL communication signals to serve the remote units 102 in the time, frequency, and / or spatial domains.

[0036] In various embodiments, the remote unit 102 and / or the network unit 104 may use the first information set to determine a parameter set, the parameter set including first information corresponding to the first model and the second model. In some embodiments, the remote unit 102 and / or the network unit 104 may send a second information set to the second device, the second information set including second information for the first model or the second model. Thus, the remote unit 102 and / or the network unit 104 may be used to determine parameters for multiple models.

[0037] In some embodiments, the remote unit 102 and / or the network unit 104 may receive an information set from a first device, the information set including first information corresponding to the first model and the second model. In some embodiments, the remote unit 102 and / or the network unit 104 may use the first information to determine a third model. In some embodiments, the remote unit 102 and / or the network unit 104 may generate an output based on the third model and the first data set. Therefore, the remote unit and / or the network unit 104 may be used to determine parameters of multiple models.

[0038] Figure 2 An embodiment of an apparatus 200 that can be used to determine parameters of multiple models is depicted. Apparatus 200 includes an embodiment of remote unit 102. In addition, remote unit 102 may include processor 202, memory 204, input device 206, display 208, transmitter 210, and receiver 212. In some embodiments, input device 206 and display 208 are combined into a single device, such as a touch screen. In some embodiments, remote unit 102 may not include any input device 206 and / or display 208. In various embodiments, remote unit 102 may include one or more of the following: processor 202, memory 204, transmitter 210, and receiver 212, and may not include input device 206 and / or display 208.

[0039] In one embodiment, the processor 202 may include any known controller capable of executing computer-readable instructions and / or capable of performing logical operations. For example, the processor 202 may be a microcontroller, a microprocessor, a central processing unit (CPU), a graphics processing unit (GPU), an auxiliary processing unit, a field programmable gate array (FPGA), or a similar programmable controller. In some embodiments, the processor 202 executes instructions stored in the memory 204 to perform the methods and routines described herein. The processor 202 is communicatively coupled to the memory 204, the input device 206, the display 208, the transmitter 210, and the receiver 212.

[0040] In one embodiment, memory 204 is a computer readable storage medium. In some embodiments, memory 204 includes volatile computer storage media. For example, memory 204 may include RAM, including dynamic RAM (DRAM), synchronous dynamic RAM (SDRAM) and / or static RAM (SRAM). In some embodiments, memory 204 includes non-volatile computer storage media. For example, memory 204 may include a hard drive, flash memory or any other suitable non-volatile computer storage device. In some embodiments, memory 204 includes both volatile and non-volatile computer storage media. In some embodiments, memory 204 also stores program code and related data, such as an operating system or other controller algorithms operating on remote unit 102.

[0041] In one embodiment, input device 206 may include any known computer input device, including a touchpad, button, keyboard, stylus, microphone, etc. In some embodiments, input device 206 may be integrated with display 208, for example, as a touch screen or similar touch-sensitive display. In some embodiments, input device 206 includes a touch screen so that text can be entered using a virtual keyboard displayed on the touch screen and / or by handwriting on the touch screen. In some embodiments, input device 206 includes two or more different devices, such as a keyboard and a touchpad.

[0042] In one embodiment, display 208 may include any known electronically controllable display or display device. Display 208 may be designed to output visual, auditory and / or tactile signals. In some embodiments, display 208 includes an electronic display capable of outputting visual data to a user. For example, display 208 may include, but is not limited to, a liquid crystal display (LCD), a light emitting diode (LED) display, an organic light emitting diode (OLED) display, a projector, or a similar display device capable of outputting images, text, etc. to a user. As another non-limiting example, display 208 may include wearable displays such as smart watches, smart glasses, and head-up displays. In addition, display 208 may be a component of a smart phone, a personal digital assistant, a television, a desktop computer, a notebook (laptop) computer, a personal computer, a vehicle dashboard, etc.

[0043] In some embodiments, the display 208 includes one or more speakers for generating sounds. For example, the display 208 can generate an audible alarm or notification (e.g., a beep or buzzer). In some embodiments, the display 208 includes one or more tactile devices for generating vibration, motion, or other tactile feedback. In some embodiments, all or part of the display 208 can be integrated with the input device 206. For example, the input device 206 and the display 208 can form a touch screen or similar touch-sensitive display. In other embodiments, the display 208 can be located near the input device 206.

[0044] In some embodiments, the processor 202 is configured to cause the apparatus to: determine a parameter set using a first information set, the parameter set comprising first information corresponding to the first model and the second model; and send a second information set to a second device, the second information set comprising second information for the first model or the second model.

[0045] In some embodiments, the processor 202 is configured to cause the apparatus to: receive an information set from a first device, the information set comprising first information corresponding to a first model and a second model; determine a third model using the first information; and generate an output based on the third model and the first data set.

[0046] Although only one transmitter 210 and one receiver 212 are shown, the remote unit 102 may have any suitable number of transmitters 210 and receivers 212. The transmitter 210 and receiver 212 may be any suitable type of transmitter and receiver. In one embodiment, the transmitter 210 and receiver 212 may be part of a transceiver.

[0047] Figure 3 One embodiment of an apparatus 300 that can be used to determine parameters of multiple models is depicted. The apparatus 300 includes one embodiment of the network unit 104. In addition, the network unit 104 can include a processor 302, a memory 304, an input device 306, a display 308, a transmitter 310, and a receiver 312. It is understood that the processor 302, the memory 304, the input device 306, the display 308, the transmitter 310, and the receiver 312 can be substantially similar to the processor 202, the memory 204, the input device 206, the display 208, the transmitter 210, and the receiver 212 of the remote unit 102, respectively.

[0048] In some embodiments, the processor 302 is configured to cause the apparatus to: determine a parameter set using a first information set, the parameter set comprising first information corresponding to the first model and the second model; and send a second information set to a second device, the second information set comprising second information for the first model or the second model.

[0049] In some embodiments, the processor 302 is configured to cause the apparatus to: receive an information set from a first device, the information set comprising first information corresponding to a first model and a second model; determine a third model using the first information; and generate an output based on the third model and the first data set.

[0050] It should be noted that one or more of the embodiments described herein may be combined into a single embodiment.

[0051] Figure 4 4 is a schematic block diagram illustrating an embodiment of a wireless network 400, the wireless network including a first UE 402 (UE-1, UE1), a second UE 404 (UE-2, UE2), a Kth UE 406 (UE-K, UE K ) and gNB 408(B1). B1 is equipped with M antennas and consists of U1, U2, ..., U K Each of the K UEs represented has N antennas. Indicates that between B1 and U k The channel at time t in the frequency band l(l∈{1,2,...,L}) between is a matrix of size N×M with complex entries, that is, .

[0052] At time t and frequency band l, gNB 408 wishes to send a signal to user U k Send Message , where k = {1, 2, …, K}, and the gNB uses As the precoding vector. k , The received signal at can be written as:

[0053]

[0054] in represents the noise vector at the receiver.

[0055] To increase the achievable rate of the link, gNB 408 selects the link that maximizes the received signal-to-noise ratio (SNR). Several different methods can be used to select , most of which rely on Some understanding of .

[0056] In some embodiments, the gNB 408 may obtain the information by direct measurement (e.g., in time domain duplex (TDD) mode and assuming reciprocity of the channel) or indirectly using information sent by the UE to the gNB 408 (e.g., in frequency division duplex (FDD) mode). In various embodiments, a large amount of feedback may be required to send information about This can be important if there are a large number of antennas or / and a large frequency band.

[0057] In some embodiments herein, only a single time slot is used, but these embodiments can be used with multiple time slots. Without loss of generality, Can be used To express.

[0058] In addition, H k can be defined as a matrix of size N×M×L, which includes the superposition of all frequency bands , for example, H k The entry at [n,m,l] is equal to . Therefore, in general, each UE needs to send information about N×M×L complex numbers to gNB 408.

[0059] In some embodiments, a two-sided model can be used to reduce the required feedback information, where the encoding part (at the UE) computes a quantized latent representation of the input data, and the decoding part (at the gNB) takes that latent representation and uses it to reconstruct the desired output.

[0060] Figure 5 1 is a schematic block diagram of an embodiment of a system 500 that uses a dual-sided model having a neural network (NN) based model on a UE side and a gNB side. The system 500 includes a UE side 502 (M e , coding model) and gNB side 504 (M d , decoding model). The UE side 502 receives input data 506 and outputs a potential representation 508. In addition, the gNB side 504 receives the potential representation 508 and outputs an output 510.

[0061] It is understood that there may be several methods for training the NN modules on the UE side 502 and the gNB side 504, including centralized training, simultaneous training, and separate training. In addition, updating the two-sided models may be performed centrally on one entity, simultaneously on different entities, or separately.

[0062] In split training and / or model updating, the NN modules of the UE part and the gNB part are trained in different training sessions (e.g., there is no forward or back-propagation path between the two parts).

[0063] In some embodiments, there may be different methods for training and / or updating the two-sided models in separate training loops. One reason for separate model training is that the UE and gNB want to use their own designed and optimized models instead of just running a model provided by another vendor.

[0064] In some embodiments, the split training of the model may start with first training the model at the UE and then training the model at the gNB side (e.g., UE first), or the training may start with first training the model at the gNB and then training the model at the UE side (e.g., gNB first). It should be noted that there may be other alternatives besides the UE first and gNB first approaches.

[0065] In the UE-first method, the UE uses the channel state information (CSI) data set D = { <x i ,o i >, i = 1, 2, ..., N} (for example, where x i is the input CSI, x i and i is the expected output) to train the local copy of the bilateral model, such as the UE part ( ) and gNB part ( )Both. The part will be used to compress x into a potential representation z. In common cases, the UE will Sent to the gNB so that it can be used as the gNB part (at the gNB), but in case of separate training, the gNB wants to use the model trained and optimized by itself.

[0066] Therefore, in one embodiment, the UE constructs a data set D u The dataset includes <z i ,o i >equal samples, where z i is the output of the encoder part, e.g., for x sampled from the CSI dataset D i , The data set is sent to the gNB. The gNB uses the data set D u To train and / or update the gNB part of the bilateral model, e.g. .

[0067] In some embodiments, not one UE, but several UEs send their data to a central location (e.g., still the same provider of UEs), and and The training of is performed using collective data. This allows the model to generalize better as more samples are observed during training time.

[0068] In the gNB-first approach, the gNB uses dataset D c , which includes CSI reports sent from one or more UEs to the gNB, such as D c ={ <x i , o i >, i = 1, 2, ..., N}. Use D c , the gNB trains a local copy of the bilateral model, e.g. the UE part ( ) and gNB part ( )Both.

[0069] The part can be used to construct the required CSI information according to the potential representation (e.g., z) fed back by the UE. i In some embodiments, the gNB will send a , so that it can be used as the UE part (e.g., at the UE), but for separate training, the UE can use the model trained and optimized by itself.

[0070] Therefore, in one embodiment, the gNB constructs a dataset D g The dataset includes <x i ,z i> Equal samples, where z i is the output of the encoder part of the gNB model, such as The gNB can: a) feedback the complete D to each UE g or b) only send the same x as the gNB receives from each specific UE i Related z i In the second alternative, the communication overhead is smaller, but the complete D g The transmission of results in better generalization of the training data. The UE then uses the received data to train and / or update the UE part of the two-sided model, e.g. .

[0071] Although the UE priority method and the gNB priority method can work, they may require high communication overhead and cause high latency. In various embodiments, there may be lower communication costs than the above-mentioned UE priority method and gNB priority method.

[0072] In some embodiments, there may be a two-sided model where each UE sends its feedback data z to the gNB. i (For example, using the collected CSI information x i and where i represents different samples). The gNB uses this data and To generate CSI data.

[0073] In another embodiment of UE-first training, the UE trains a local copy (e.g., UE part ( ) and gNB part ( ) both). Afterwards, the gNB part of the model trained at the UE (or multiple UEs) is sent to the gNB. If you need to reduce communication overhead, you can train so that it has low resolution NN weights.

[0074] In the UE-first training scheme, the gNB can receive a set of <z i ,o i > to train the model.

[0075] In some embodiments, the UE may to the gNB, and then only sends a set of z to the gNB i The gNB can then use Locally i It can then be estimated using To construct the training information required for the training of the gNB part of the bilateral model, for example It should be noted that due to o i is not a quantitative representation, so it is the same as Its transmission may result in higher overhead in the communication system compared to the transmission of

[0076] For model monitoring and model updating, it is assumed that there is a trained version of and Available and running at UE and gNB respectively. In this solution, Trained to imitate Therefore, since the UE has and Both, it can periodically perform model monitoring locally. For example, it can compute a dissimilarity measure between the expected output and the gNB output estimate (e.g., using replace ). For example, it can be used If the dissimilarity becomes larger than a threshold, it can initiate an update procedure or can signal the gNB that the model needs to be updated.

[0077] After the update process is initiated, since the UE can access the newly collected CSI data, such as x i , o i , the UE can use them together with the initial training data to update the local model and After the model is updated at the UE, it may send additional training data to the gNB, such as a set of Alternatively, it can only update With the newly collected CSI z i This enables the gNB to construct o i Without its direct transmission, The resulting dataset can be used to update the gNB It should be noted that, in contrast to the initial training phase, in the update phase, no o i (e.g., due to its potentially high communication overhead) may be more important.

[0078] In another embodiment of gNB-first training, the gNB first trains a local copy of the two-sided model, e.g., the UE part ( ) and gNB part ( ) Both. The gNB part of the model trained at the gNB is sent to the UE. If you need to reduce communication overhead, you can train so that it has low resolution NN weights.

[0079] For initial training in the gNB priority scheme, in order to train , UE needs to receive a set of z i (For example, corresponding to x previously sent by the UE to the gNB i ) or a new set <x u ,z i >.

[0080] In some embodiments, the gNB may sent to UE without sending z i In fact, with UE can construct a local two-sided model Come to train , where it will The weights of x are kept fixed and only the CSI data collected from the UE is used (e.g., x i )train For model monitoring and model updating, it can be assumed that there is a trained version of and Available and running at UE and gNB respectively. and Both, so that model monitoring can be performed periodically. For example, it can be done using To calculate the dissimilarity measure between the expected output and the model output present at the gNB. If the dissimilarity becomes larger than a threshold, it can initiate an update process or can signal the gNB that the model needs to be updated.

[0081] In some embodiments, for initiation of the update process, the UE may first attempt to update its encoder network , and checks whether it can solve the dissimilarity problem. To do this, it can construct a local two-sided model , where it will The weights of x are kept fixed and only the CSI data collected from the UE is used (e.g., x i )train If successful, the UE uses the new , while gNB uses the original .if If the local update fails to improve the performance, the UE sends new training data to the gNB to start gNB-priority training, or it can switch to UE-priority training to update the model.

[0082] In various embodiments, there may be The UE part of the model trained at the gNB is sent to the UE. It should be noted that if necessary, in order to reduce communication overhead, so that it has low resolution NN weights.

[0083] In the gNB priority scheme, in order to train , UE needs to receive a set of z i (For example, corresponding to x previously sent by the UE to the gNB i ) or a new set <x i ,z i >.

[0084] In some embodiments, the gNB may sent to UE without sending z i In fact, with , when collecting CSI information (x i ) After that, the UE can generate z i ,Right now, The obtained <x i ,z i >Datasets can be used for training .

[0085] In various embodiments, for model updating, it may be assumed that there is a trained version of and Available and running at UE and gNB respectively. If a model update is required, the gNB can first instruct the UE to send new training data <x i ,o i > so that it can be updated and . With new , gNB can Sent to UE so that it can generate z i ,Right now, , and then use the resulting <x i ,z i >Dataset to train It should be noted that for all x i ,feedback The cost of feedback z i , and it can also be used for newly observed x i , in order to obtain a model with better generalization ability.

[0086] Figure 6 is a flow chart illustrating one embodiment of a method 600 for determining parameters of a plurality of models. In some embodiments, the method 600 is performed by a device such as the remote unit 102 and / or the network unit 104. In certain embodiments, the method 600 may be performed by a processor executing program code, such as a microcontroller, a microprocessor, a CPU, a GPU, an auxiliary processing unit, an FPGA, etc.

[0087] In various embodiments, the method 600 includes determining 602 a parameter set at a first device using a first information set, the parameter set including first information corresponding to a first model and a second model. In some embodiments, the method 600 includes sending 604 a second information set to a second device, the second information set including second information for the first model or the second model.

[0088] In some embodiments, the first device comprises a UE and the second device comprises a network device. In some embodiments, the first information set comprises input data and expected output data of the two-part model. In various embodiments, the input data and the expected output data are related to channel state information.

[0089] In one embodiment, the first model and the second model are used to determine a potential representation of the input data and to generate expected output data based on the potential representation. In some embodiments, the second information set includes characterization information of the second model. In some embodiments, method 600 also includes determining the first data based on the first model and the input channel data.

[0090] In various embodiments, the representation of the first data is sent to the second device. In one embodiment, the method 600 also includes determining whether to update the parameter set based on the first model, the second model, or a combination thereof. In some embodiments, the method 600 also includes sending an update request to the second device.

[0091] In some embodiments, the method 600 further comprises determining an updated parameter set based on a third information set, wherein the third information set comprises input data and expected output data of the two-part model. In various embodiments, the method 600 further comprises sending an update to the second information set based on the updated parameter set. In one embodiment, the first device comprises a network device, and the second device comprises a UE.

[0092] In some embodiments, the first information set is received from the second device. In some embodiments, the first information set includes input data and expected output data of the two-part model. In various embodiments, the first model and the second model are used to determine a potential representation of the input data and generate the expected output data based on the potential representation.

[0093] In one embodiment, the second information set includes characterization information of the second model. In some embodiments, the second information set includes characterization information of the first model.

[0094] In some embodiments, the network device comprises a next gNB. In various embodiments, the first model and the second model comprise a limited bit weight resolution.

[0095] Figure 7 is a flow chart illustrating another embodiment of a method 700 for determining parameters of a plurality of models. In some embodiments, the method 700 is performed by a device such as the remote unit 102 and / or the network unit 104. In some embodiments, the method 700 may be performed by a processor executing program code, such as a microcontroller, a microprocessor, a CPU, a GPU, an auxiliary processing unit, an FPGA, etc.

[0096] In various embodiments, the method 700 includes receiving 702 a set of information at a second device from a first device, the set of information including first information corresponding to a first model and a second model. In some embodiments, the method 700 includes determining 704 a third model using the first information. In certain embodiments, the method 700 includes generating 706 an output based on the third model and the first data set.

[0097] In some embodiments, the first device comprises a UE and the second device comprises a network device. In some embodiments, the first data set is received from the first device. In various embodiments, the output is determined based on a third model.

[0098] In one embodiment, the first model and the second model are used to determine a potential representation of the input data and generate expected output data based on the potential representation. In some embodiments, the information set includes characterization information of the second model. In some embodiments, the second information set includes characterization information of the first model.

[0099] In various embodiments, the first device comprises a network device and the second device comprises a UE. In one embodiment, the first data set is based on channel data. In some embodiments, the method 700 further comprises sending the output to the first device.

[0100] In some embodiments, the first model and the second model are used to determine a potential representation of the input data and generate expected output data based on the potential representation. In various embodiments, the information set includes characterization information of the second model. In one embodiment, method 700 also includes determining whether to update the parameter set based on the third model and the first information set.

[0101] In some embodiments, the method 700 further comprises sending an update request to the first device. In some embodiments, the method 700 further comprises receiving the update request from the first device. In various embodiments, the method 700 further comprises sending a second data set to the first device, wherein the second data set is based on the channel data.

[0102] In one embodiment, the method 700 further comprises receiving an updated set of information from the first device. In some embodiments, the set of information comprises characterization information of the first model. In some embodiments, the network device comprises a next gNB.

[0103] In various embodiments, determining the third model includes initial training of a NN parameter set for the third model. In one embodiment, determining the third model includes updating the NN parameter set for the third model.

[0104] In one embodiment, an apparatus for wireless communication comprises: a processor; and a memory coupled to the processor, the processor being configured to cause the apparatus to: determine a parameter set using a first information set, the parameter set comprising first information corresponding to a first model and a second model; and send a second information set to a second device, the second information set comprising second information for the first model or the second model.

[0105] In certain embodiments, the apparatus comprises a UE, and the second device comprises a network device.

[0106] In some embodiments, the first set of information includes input data and expected output data of the two-part model.

[0107] In various embodiments, the input data and the expected output data are correlated with channel state information.

[0108] In one embodiment, the first model and the second model are used to determine a latent representation of the input data and to generate expected output data based on the latent representation.

[0109] In some embodiments, the second set of information includes characterization information of the second model.

[0110] In some embodiments, the processor is further configured to cause the apparatus to determine first data based on the first model and the input channel data.

[0111] In various embodiments, a representation of the first data is sent to a second device.

[0112] In one embodiment, the processor is further configured to cause the apparatus to determine whether to update the parameter set based on the first model, the second model, or a combination thereof.

[0113] In certain embodiments, the processor is further configured to cause the apparatus to send an update request to the second device.

[0114] In some embodiments, the processor is further configured to cause the apparatus to determine an updated set of parameters based on a third set of information, and the third set of information includes input data and expected output data of the two-part model.

[0115] In various embodiments, the processor is further configured to cause the apparatus to send an update to the second set of information based on the updated set of parameters.

[0116] In one embodiment, the apparatus comprises a network device, and the second device comprises a UE.

[0117] In some embodiments, the first set of information is received from a second device.

[0118] In some embodiments, the first set of information includes input data and expected output data of the two-part model.

[0119] In various embodiments, the first model and the second model are used to determine a latent representation of the input data and to generate expected output data based on the latent representation.

[0120] In one embodiment, the second information set includes characterization information of the second model.

[0121] In some embodiments, the second set of information includes characterization information of the first model.

[0122] In some embodiments, the network device includes a next gNB.

[0123] In various embodiments, the first model and the second model include a finite bit weight resolution.

[0124] In one embodiment, a method for wireless communication at a first device includes: using a first information set to determine a parameter set, the parameter set including first information corresponding to a first model and a second model; and sending a second information set to a second device, the second information set including second information for the first model or the second model.

[0125] In certain embodiments, the first device comprises a UE and the second device comprises a network device.

[0126] In some embodiments, the first set of information includes input data and expected output data of the two-part model.

[0127] In various embodiments, the input data and the expected output data are correlated with channel state information.

[0128] In one embodiment, the first model and the second model are used to determine a latent representation of the input data and to generate expected output data based on the latent representation.

[0129] In some embodiments, the second set of information includes characterization information of the second model.

[0130] In some embodiments, the method further comprises determining first data based on the first model and the input channel data.

[0131] In various embodiments, a representation of the first data is sent to a second device.

[0132] In one embodiment, the method further includes determining whether to update the parameter set based on the first model, the second model, or a combination thereof.

[0133] In some embodiments, the method further includes sending an update request to the second device.

[0134] In some embodiments, the method further comprises determining an updated set of parameters based on a third set of information, wherein the third set of information comprises input data and expected output data of the two-part model.

[0135] In various embodiments, the method further includes sending an update to the second set of information based on the updated set of parameters.

[0136] In one embodiment, the first device comprises a network device and the second device comprises a UE.

[0137] In some embodiments, the first set of information is received from a second device.

[0138] In some embodiments, the first set of information includes input data and expected output data of the two-part model.

[0139] In various embodiments, the first model and the second model are used to determine a latent representation of the input data and to generate expected output data based on the latent representation.

[0140] In one embodiment, the second information set includes characterization information of the second model.

[0141] In some embodiments, the second set of information includes characterization information of the first model.

[0142] In some embodiments, the network device includes a next gNB.

[0143] In various embodiments, the first model and the second model include a finite bit weight resolution.

[0144] In one embodiment, an apparatus for wireless communication includes: a processor; and a memory coupled to the processor, the processor being configured to cause the apparatus to: receive an information set from a first device, the information set including first information corresponding to a first model and a second model; determine a third model using the first information; and generate an output based on the third model and the first data set.

[0145] In certain embodiments, the first device comprises a UE, and the apparatus comprises a network device.

[0146] In some embodiments, the first data set is received from a first device.

[0147] In various embodiments, the output is determined based on a third model.

[0148] In one embodiment, the first model and the second model are used to determine a latent representation of the input data and generate expected output data based on the latent representation.

[0149] In some embodiments, the set of information includes characterization information of the second model.

[0150] In some embodiments, the second set of information includes characterization information of the first model.

[0151] In various embodiments, the first device comprises a network device, and the apparatus comprises a UE.

[0152] In one embodiment, the first data set is based on channel data.

[0153] In certain embodiments, the processor is further configured to cause the apparatus to send the output to the first device.

[0154] In some embodiments, the first model and the second model are used to determine a latent representation of the input data and to generate expected output data based on the latent representation.

[0155] In various embodiments, the set of information includes characterization information of the second model.

[0156] In one embodiment, the processor is further configured to cause the apparatus to determine whether to update the parameter set based on the third model and the first information set.

[0157] In certain embodiments, the processor is further configured to cause the apparatus to send an update request to the first device.

[0158] In some embodiments, the processor is further configured to cause the apparatus to receive an update request from the first device.

[0159] In various embodiments, the processor is further configured to cause the apparatus to send a second data set to the first device, wherein the second data set is based on the channel data.

[0160] In one embodiment, the processor is further configured to cause the apparatus to receive an updated set of information from the first device.

[0161] In some embodiments, the set of information includes characterization information of the first model.

[0162] In some embodiments, the network device includes a next gNB.

[0163] In various embodiments, the processor is configured to cause the device to determine the third model, including the processor being further configured to cause the device to initially train a set of NN parameters for the third model.

[0164] In one embodiment, the processor is configured to cause the device to determine the third model, including the processor being further configured to cause the device to update a NN parameter set of the third model.

[0165] In one embodiment, a method for wireless communication at a second device includes: receiving an information set from a first device, the information set including first information corresponding to a first model and a second model; determining a third model using the first information; and generating an output based on the third model and the first data set.

[0166] In certain embodiments, the first device comprises a UE and the second device comprises a network device.

[0167] In some embodiments, the first data set is received from a first device.

[0168] In various embodiments, the output is determined based on a third model.

[0169] In one embodiment, the first model and the second model are used to determine a latent representation of the input data and to generate expected output data based on the latent representation.

[0170] In some embodiments, the set of information includes characterization information of the second model.

[0171] In some embodiments, the second set of information includes characterization information of the first model.

[0172] In various embodiments, the first device comprises a network device and the second device comprises a UE.

[0173] In one embodiment, the first data set is based on channel data.

[0174] In some embodiments, the method further includes sending the output to the first device.

[0175] In some embodiments, the first model and the second model are used to determine a latent representation of the input data and to generate expected output data based on the latent representation.

[0176] In various embodiments, the set of information includes characterization information of the second model.

[0177] In one embodiment, the method further comprises determining whether to update the parameter set based on the third model and the first set of information.

[0178] In some embodiments, the method further includes sending an update request to the first device.

[0179] In some embodiments, the method further includes receiving an update request from the first device.

[0180] In various embodiments, the method further includes sending a second data set to the first device, wherein the second data set is based on the channel data.

[0181] In one embodiment, the method further comprises receiving an updated set of information from the first device.

[0182] In some embodiments, the set of information includes characterization information of the first model.

[0183] In some embodiments, the network device includes a next gNB.

[0184] In various embodiments, determining the third model includes initial training of a set of NN parameters for the third model.

[0185] In one embodiment, determining the third model includes updating a set of NN parameters of the third model.

[0186] Embodiments may be practiced in other specific forms. The described embodiments are considered in all respects to be illustrative only, and not restrictive. Therefore, the scope of the present invention is indicated by the appended claims rather than by the above description. All changes within the equivalent meaning and scope of the claims should be included within their scope.

Claims

1. A user equipment UE, comprising: at least one memory; as well as at least one processor coupled to the at least one memory and configured to cause the UE to: determining a parameter set using the first information set, the parameter set including first information corresponding to the first model and the second model; as well as A second information set is sent to a second device, where the second information set includes second information for the first model or the second model. The UE according to claim 1 , wherein the second device comprises a network device.

3. The UE according to claim 2, wherein the first information set includes input data and expected output data of the two-part model. The UE according to claim 3 , wherein the input data and the expected output data are related to channel state information. 5 . The UE of claim 3 , wherein the first model and the second model are used to determine a potential representation of the input data and to generate the expected output data based on the potential representation. The UE according to claim 2 , wherein the second information set includes characterization information for the second model. 7 . The UE of claim 2 , wherein the at least one processor is configured to cause the UE to: determine first data based on the first model and input channel data.

8. The UE of claim 7, wherein the representation of the first data is sent to the second device.

9. The UE of claim 2, wherein the at least one processor is configured to cause the UE to: determine whether to update the parameter set based on the first model, the second model, or a combination thereof. 10 . The UE of claim 9 , wherein the at least one processor is configured to cause the UE to: send an update request to the second device.

11. The UE of claim 9, wherein the at least one processor is configured to cause the UE to: determine an updated parameter set based on a third information set, and the third information set includes input data and expected output data of the two-part model. 12 . The UE of claim 11 , wherein the at least one processor is configured to cause the UE to: send an update to the second set of information based on the updated set of parameters.

13. The UE of claim 1, wherein the first information set comprises input data and expected output data of a two-part model.

14. The UE of claim 13, wherein the first model and the second model are used to determine a potential representation of the input data and generate the expected output data based on the potential representation.

15. The UE of claim 1, wherein the second information set comprises characterization information for the second model.

16. A processor for wireless communication, comprising: at least one controller coupled to the at least one memory and configured to cause the processor to: determining a parameter set using the first information set, the parameter set including first information corresponding to the first model and the second model; as well as A second information set is sent to a second device, where the second information set includes second information for the first model or the second model.

17. A method performed by a user equipment UE, the method comprising: determining a parameter set using the first information set, the parameter set including first information corresponding to the first model and the second model; as well as A second information set is sent to a second device, where the second information set includes second information for the first model or the second model.

18. An apparatus for performing a network function, the apparatus comprising: at least one memory; as well as at least one processor coupled to the at least one memory and configured to cause the apparatus to: receiving, from a first device, an information set including first information corresponding to a first model and a second model; using the first information to determine a third model; and An output is generated based on the third model and the first data set.

19. The apparatus of claim 18, wherein the first device comprises a user equipment (UE), and the apparatus comprises a network device.

20. The apparatus of claim 19, wherein the first data set is received from the first device.

Images