Neural network model training method and communication apparatus

Abstract

The present application provides a neural network model training method and a communication apparatus. The method comprises: a sending end device acquires a target feedback mode and a target update mode; at a feedback time corresponding to the target feedback mode, receives a feedback result for a forward reasoning result from a receiving end device, wherein the forward reasoning result is obtained on the basis of a first neural network model deployed in the sending end device; and further, the sending end device updates model parameters of the first neural network model on the basis of the target update mode and the feedback result. By means of such a neural network model training method, the sending end device can know the time when the receiving end device sends the feedback result, and know the update mode, thereby improving the transparency of the training duration of a neural network model to the sending end device, and optimizing the performance of a wireless communication system applying the neural network model.

Description

A neural network model training method and communication device Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a neural network model training method and a communication device. Background Art

[0002] Combining wireless communication systems with neural networks—that is, training the neural network models within wireless communication systems through data-driven training—helps improve the performance of wireless communication systems. Neural network models can be applied to various devices in wireless communication systems (including receivers and transmitters). For example, applying a neural network model to the channel decoding module in a receiver device through data-driven training can help improve the channel decoding performance of that receiver device.

[0003] A training method for updating a neural network model deployed on a device (including a transmitting device or a receiving device) using training data from a communication network is beneficial for improving the compatibility of the neural network model with the communication channel, thereby improving the performance of the wireless communication system that uses the neural network model. However, during such a training process, the transmitting device or the receiving device may still be performing other services, resulting in the resources (including but not limited to computing resources, storage space, etc.) available for neural network model training on the transmitting device or the receiving device being unstable, thereby causing the training duration of the neural network model to be unstable, affecting the performance of the wireless communication system that uses the neural network model.

[0004] Summary of the Invention

[0005] The present application provides a neural network model training method and communication device. During the training process, the sending device can know the time when the receiving device sends the feedback result, which is beneficial for the sending device to know the training duration during the neural network model training process, thereby facilitating the optimization of the performance of the wireless communication system that applies the neural network model.

[0006] In the first aspect, the present application provides a neural network model training method, which is applied to a sending device, or to a module in the sending device (such as a chip or a chip system, etc.). Taking the application to the sending device as an example, the method includes: the sending device obtains a target feedback mode and a target update mode; further, the sending device receives a feedback result for a forward reasoning result from the receiving device at a feedback time corresponding to the target feedback mode, and the forward reasoning result is obtained based on a first neural network model deployed in the sending device; and based on the target update mode and the feedback result, the model parameters of the first neural network model are updated.

[0007] Based on the method described in the first aspect, after the first neural network model is deployed on the transmitting device, before the transmitting device trains the first neural network model, the transmitting device can first know the target feedback mode of the receiving device for sending feedback results, and know the update mode of the neural network model deployed on the transmitting device. This is conducive to reducing the unknowns of the transmitting device about the training duration of the neural network model during training, that is, the transmitting device knows the feedback time of the feedback result, which is conducive to the transmitting device determining the training duration of the neural network model, thereby facilitating the optimization of the performance of the wireless communication system that applies the neural network model.

[0008] In one possible implementation, a transmitting device obtains one or more of the transmitting device's first capability information, the receiving device's second capability information, or first channel information, where the first channel information is information corresponding to the channel between the transmitting device and the receiving device. Furthermore, the transmitting device determines a target feedback mode based on one or more of the first capability information, the second capability information, or the first channel information. Furthermore, the transmitting device determines a target update mode based on one or more of the first capability information, the second capability information, or the first channel information. This possible implementation, in which the transmitting device determines the target feedback mode and the target update mode, facilitates improving the timeliness with which the transmitting device learns the target feedback mode and the target update mode.

[0009] In one possible implementation, the target feedback mode is the first feedback mode, and the transmitting device determines the feedback time based on the time interval and the sending time of the forward reasoning result; wherein the time interval is the interval between the sending time of the forward reasoning result and the feedback time of the feedback result, and the time interval is determined based on one or more of the first capability information, the second capability information, or the first channel information. By implementing this possible implementation, when the target feedback mode is the first feedback mode, by determining the time interval based on one or more of the first capability information, the second capability information, or the first channel information, and then determining the feedback time based on the time interval, it is beneficial to improve the adaptability of the feedback time to the actual communication application scenario (i.e., the scenario involving the transmitting device, the receiving device, and the channel between the transmitting device and the second notification device).

[0010] In one possible implementation, the target feedback mode is the second feedback mode, and the transmitting device receives first indication information from the receiving device, where the first indication information is used to indicate a feedback time. By implementing this possible implementation, when the target feedback mode is the second feedback mode, the receiving device indicates the feedback time to the transmitting device. This facilitates determining the feedback time based on the specific circumstances of the receiving device (e.g., the current capabilities of the receiving device), thereby improving the adaptability of the feedback time to actual communication application scenarios.

[0011] In one possible implementation, the target feedback mode is the second feedback mode. The transmitting device receives a request message from the receiving device requesting the transmission of feedback results. Furthermore, the transmitting device sends a confirmation message to the receiving device, the confirmation message including first indication information indicating a feedback time. By implementing this possible implementation, when the target feedback mode is the second feedback mode, if the receiving device determines that feedback results can be sent, the receiving device requests the transmitting device to send the feedback results, and the transmitting device determines the feedback time. This helps improve the adaptability of the feedback time to actual communication application scenarios.

[0012] In one possible implementation, there are multiple forward reasoning results and multiple feedback results, and there is a one-to-one correspondence between the forward reasoning results and the feedback results. The transmitting device receives second indication information from the receiving device, where the second indication information is used to indicate the correspondence between each feedback result and each forward reasoning result. By implementing this possible implementation, when there are multiple feedback results, the receiving device indicates the correspondence between each feedback result and the forward reasoning result, thereby improving the efficiency of the transmitting device in determining the correspondence between the feedback results and the forward reasoning results.

[0013] In one possible implementation, the target update mode is a first update mode or a second update mode; wherein the first update mode indicates an update after receiving N feedback results, where N is an integer greater than or equal to 1; the second update mode indicates an update after receiving third indication information, where the third indication information is used to indicate an update of model parameters of the first neural network model. By implementing this possible implementation, the target update mode can be any of a variety of modes, which is conducive to improving the flexibility of the neural network model training process.

[0014] In the second aspect, the present application provides a neural network model training method, which is applied to a receiving device, or to a module in the receiving device (such as a chip or chip system, etc.). Taking the application to the receiving device as an example, the method includes: the receiving device obtains a target feedback mode; and at the feedback time corresponding to the target feedback mode, sends a feedback result for the forward reasoning result to the sending device; the forward reasoning result is obtained based on a first neural network model deployed in the sending device.

[0015] Based on the beneficial effects of the method described in the second aspect, reference may be made to the beneficial effects of the implementation described in the first aspect, which will not be repeated hereafter.

[0016] In one possible implementation, the receiving device obtains one or more of the first capability information of the sending device, the second capability information of the receiving device, or the first channel information, where the first channel information is information corresponding to the channel between the sending device and the receiving device; the receiving device determines the target feedback mode based on one or more of the first capability information, the second capability information, or the first channel information.

[0017] In one possible implementation, the target feedback mode is the first feedback mode, and the receiving device determines the feedback time based on the time interval and the sending time of the forward reasoning result; wherein the time interval is the interval between the sending time of the forward reasoning result and the feedback time of the feedback result, and the time interval is determined based on one or more of the first capability information, the second capability information, or the first channel information.

[0018] In a possible implementation, the target feedback mode is the second feedback mode, and the receiving end device sends first indication information to the transmitting end device, where the first indication information is used to indicate a feedback time.

[0019] In one possible implementation, the target feedback mode is the second feedback mode, and the receiving device sends a request message to the sending device, where the request message is used to request sending a feedback result; the receiving device receives a confirmation message from the sending device, where the confirmation message includes first indication information, where the first indication information is used to indicate a feedback time.

[0020] In one possible implementation, there are multiple forward reasoning results and multiple feedback results, and the forward reasoning results and the feedback results correspond one to one. The receiving device sends a second indication message to the sending device, and the second indication message is used to indicate the correspondence between each feedback result and each forward reasoning result.

[0021] In one possible implementation, the receiving device obtains a target update mode, which is a first update mode or a second update mode; wherein the first update mode indicates updating after receiving N feedback results, where N is an integer greater than or equal to 1; and the second update mode indicates updating after receiving third indication information, where the third indication information is used to indicate updating of model parameters of the first neural network model.

[0022] In a possible implementation manner, the target update mode is the second update mode, and the receiving device sends third indication information to the sending device.

[0023] In a possible implementation, the receiving device is deployed with a second neural network model, and the receiving device updates the model parameters of the second neural network model according to the forward reasoning result.

[0024] In a third aspect, the present application provides a communication device, which may be a transmitting end device, a device in a transmitting end device, or a device that can be used in conjunction with a transmitting end device. The communication device may also be a chip system. The communication device may execute the method described in the first aspect. The functions of the communication device may be implemented by hardware, or by hardware executing corresponding software implementations. The hardware or software includes one or more units or modules corresponding to the above functions. The units or modules may be software and / or hardware. The operations and beneficial effects performed by the communication device may refer to the method and beneficial effects described in the first aspect above.

[0025] In a fourth aspect, the present application provides a communication device, which may be a receiving device, a device in a receiving device, or a device that can be used in conjunction with a receiving device. The communication device may also be a chip system. The communication device may execute the method described in the second aspect. The functions of the communication device may be implemented by hardware or by hardware executing corresponding software. The hardware or software includes one or more units or modules corresponding to the above functions. The units or modules may be software and / or hardware. The operations and beneficial effects performed by the communication device may refer to the method and beneficial effects described in the second aspect above.

[0026] In a fifth aspect, the present application provides a communication device, which includes a processor and an interface circuit, wherein the interface circuit is used to receive signals from other communication devices outside the communication device and transmit them to the processor or send signals from the processor to other communication devices outside the communication device, and the processor is used to implement the method as described in the first aspect through a logic circuit or executing code instructions, or the processor is used to implement the method as described in the second aspect through a logic circuit or executing code instructions.

[0027] In a sixth aspect, the present application provides a computer-readable storage medium, in which a computer program or instruction is stored. When the computer program or instruction is executed by a communication device, the method described in the first aspect or the method described in the second aspect is implemented.

[0028] In a seventh aspect, the present application provides a computer program product comprising instructions, which, when a communication device reads and executes the instructions, causes the communication device to execute the method as described in the first aspect, or causes the communication device to execute the method as described in the second aspect.

[0029] In an eighth aspect, the present application provides a communication system comprising a communication device for executing the method described in the first aspect and a communication device for executing the method described in the second aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] FIG1a is a schematic diagram of the architecture of a communication system provided in an embodiment of the present application;

[0031] FIG1b is another schematic diagram of a wireless communication system applicable to an embodiment of the present application;

[0032] FIG2 is a schematic diagram of a wireless communication system based on a neural network transmitter and receiver provided in an embodiment of the present application;

[0033] FIG3 is a schematic diagram of different modules in a transceiver optimized by several neural network models provided in an embodiment of the present application;

[0034] FIG4 is a schematic diagram of a training process of an intelligent air interface provided in an embodiment of the present application;

[0035] FIG5 is a flow chart of a neural network model training method provided in an embodiment of the present application;

[0036] FIG6 is a schematic diagram of a feedback mode provided in an embodiment of the present application;

[0037] FIG7 is a schematic diagram of another feedback mode provided in an embodiment of the present application;

[0038] FIG8 is a schematic diagram of updating a first neural network model provided by an embodiment of the present application;

[0039] FIG9 is a schematic diagram of another method for updating the first neural network model according to an embodiment of the present application;

[0040] FIG10 is a schematic diagram of another method of updating the first neural network model provided in an embodiment of the present application;

[0041] FIG11 is a schematic structural diagram of a communication device provided in an embodiment of the present application;

[0042] FIG12 is a schematic structural diagram of another communication device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solutions and advantages of this application clearer, this application will be further described in detail below with reference to the accompanying drawings.

[0044] The terms "first" and "second" and the like in the specification, claims, and drawings of this application are used to distinguish between different objects, not to describe a particular order. Furthermore, the terms "including" and "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of operations or units is not limited to the listed operations or units, but may optionally include operations or units not listed, or may optionally include other operations or units inherent to the process, method, product, or apparatus.

[0045] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0046] In order to facilitate a detailed understanding of the embodiments of the present application, the system architecture involved in the embodiments of the present application is first introduced below.

[0047] Figure 1a is a schematic diagram of the architecture of a communication system 1000 used in an embodiment of the present application. As shown in Figure 1a, the communication system includes a radio access network (RAN) 100 and a core network 200. Optionally, the communication system 1000 may also include the Internet 300. The RAN 100 includes at least one RAN node (such as 110a and 110b in Figure 1a, collectively referred to as 110) and may also include at least one terminal (such as 120a-120j in Figure 1a, collectively referred to as 120). The RAN 100 may also include other RAN nodes, such as wireless relay devices and / or wireless backhaul devices (not shown in Figure 1a). The terminal 120 is wirelessly connected to the RAN node 110, and the RAN node 110 is wirelessly or wiredly connected to the core network 200. The core network devices in the core network 200 and the RAN node 110 in the RAN 100 may be independent and different physical devices, or they may be the same physical device that integrates the logical functions of the core network devices and the logical functions of the RAN nodes. Terminals and RAN nodes may be connected to each other via wired or wireless means.

[0048] RAN100 may be an evolved universal terrestrial radio access (E-UTRA) system, a new radio (NR) system, or a future radio access system defined in the 3rd Generation Partnership Project (3GPP). RAN100 may also include two or more of the aforementioned different radio access systems. RAN100 may also be an open RAN (O-RAN).

[0049] A RAN node, also known as a radio access network device, RAN entity, or access node, and hereinafter referred to as a network device, facilitates wireless access to a communication system by a terminal. In one application scenario, a RAN node can be a base station, an evolved NodeB (eNodeB), a transmission reception point (TRP), a next-generation NodeB (gNB) in a fifth-generation (5G) mobile communication system, a next-generation base station in a sixth-generation (6G) mobile communication system, or a base station in a future mobile communication system. A RAN node can be a macro base station (such as 110a in Figure 1a), a micro base station, an indoor station (such as 110b in Figure 1a), a relay node, or a donor node.

[0050] In another application scenario, multiple RAN nodes can collaborate to help terminals achieve wireless access, with different RAN nodes implementing portions of the base station's functions. For example, a RAN node can be a centralized unit (CU), a distributed unit (DU), or a radio unit (RU). The CU implements the base station's radio resource control protocol and packet data convergence protocol (PDCP) functions, as well as the service data adaptation protocol (SDAP) functions. The DU implements the base station's radio link control layer and medium access control (MAC) layer functions, as well as some or all of the physical layer functions. For detailed descriptions of each of the above protocol layers, please refer to the relevant 3GPP technical specifications. The RU can be used to implement the transmission and reception of radio frequency signals. The CU and DU can be two independent RAN nodes, or they can be integrated into the same RAN node, such as in a baseband unit (BBU). The RU can be included in radio frequency equipment, such as a remote radio unit (RRU) or an active antenna unit (AAU). The CU can be further divided into two types of RAN nodes: CU-control plane and CU-user plane.

[0051] The RAN node can support one or more types of fronthaul interfaces. Different fronthaul interfaces correspond to DUs and RUs with different functions. If the fronthaul interface between the DU and the RU is the common public radio interface (CPRI), the DU is configured to implement one or more baseband functions, and the RU is configured to implement one or more radio frequency functions. If the fronthaul interface between the DU and the RU is the enhanced common public radio interface (eCPRI), relative to CPRI, part of the downlink and / or uplink baseband functions are moved from the DU to the RU for implementation. The division between the DU and the RU is different, corresponding to different types (category, Cat) of eCPRI, such as eCPRI Cat A, B, C, D, E, and F.

[0052] Taking eCPRI Cat A as an example, for downlink transmission, based on layer mapping, the DU is configured to implement layer mapping and one or more functions preceding it (i.e., one or more of coding, rate matching, scrambling, modulation, and layer mapping). Other functions after layer mapping (e.g., RE mapping, digital beamforming (BF), or one or more of inverse fast Fourier transform (IFFT) / cyclic prefix (CP) addition) are moved to the RU for implementation. For uplink transmission, based on RE demapping, the DU is configured to implement demapping and one or more functions preceding it (i.e., one or more of decoding, rate matching, descrambling, demodulation, inverse discrete Fourier transform (IDFT), channel equalization, and RE demapping). Other functions after demapping (e.g., one or more of digital BF or fast Fourier transform (FFT) / CP removal) are moved to the RU for implementation. It is understandable that for the functional description of DU and RU corresponding to various types of eCPRI, reference can be made to the eCPRI protocol, which will not be described in detail here.

[0053] In one possible design, the processing unit for implementing baseband functions in the BBU is called a baseband high layer (BBH) unit, and the processing unit for implementing baseband functions in the RRU / AAU / RRH is called a baseband low layer (BBL) unit.

[0054] In different systems, RAN nodes may have different names. For example, in an O-RAN system, a CU may be called an open CU (O-CU), a DU may be called an open DU (O-DU), and a RU may be called an open RU (O-RU). The RAN node in the embodiments of the present application may be implemented by a software module, a hardware module, or a combination of a software module and a hardware module. For example, the RAN node may be a server loaded with the corresponding software module. The embodiments of the present application do not limit the specific technology and specific device form adopted by the RAN node. For ease of description, the following description takes a base station as an example of a RAN node.

[0055] A terminal is a device with wireless transceiver capabilities that can send signals to a base station or receive signals from a base station. A terminal may also be referred to as a terminal device, user equipment (UE), mobile station, mobile terminal, etc. Terminals can be widely used in various scenarios, such as device-to-device (D2D), vehicle-to-everything (V2X) communication, machine-type communication (MTC), Internet of Things (IoT), virtual reality, augmented reality, industrial control, autonomous driving, telemedicine, smart grid, smart furniture, smart office, smart wearable, smart transportation, smart city, etc. A terminal can be a mobile phone, tablet computer, computer with wireless transceiver capabilities, wearable device, vehicle, airplane, ship, robot, robotic arm, smart home device, etc. The embodiments of this application do not limit the specific technology and specific device form adopted by the terminal.

[0056] Base stations and terminals can be fixed or mobile. They can be deployed on land, indoors or outdoors, handheld or vehicle-mounted; on water; or on aircraft, balloons, and satellites. The embodiments of this application do not limit the application scenarios of base stations and terminals.

[0057] The roles of base stations and terminals can be relative. For example, the helicopter or drone 120i in Figure 1a can be configured as a mobile base station. To terminals 120j accessing the wireless access network 100 via 120i, terminal 120i is a base station. However, to base station 110a, 120i is a terminal, meaning that communication between 110a and 120i occurs via a wireless air interface protocol. Of course, communication between 110a and 120i can also occur via a base station-to-base station interface protocol. In this case, 120i is also a base station relative to 110a. Therefore, base stations and terminals can be collectively referred to as communication devices. 110a and 110b in Figure 1a can be referred to as communication devices with base station functionality, while 120a-120j in Figure 1a can be referred to as communication devices with terminal functionality.

[0058] Communication between base stations and terminals, between base stations, and between terminals can be carried out through authorized spectrum, unauthorized spectrum, or both; communication can be carried out through spectrum below 6 gigahertz (GHz), spectrum above 6 GHz, or spectrum below 6 GHz and spectrum above 6 GHz. The embodiments of the present application do not limit the spectrum resources used for wireless communication.

[0059] In the embodiments of the present application, the functions of the base station may also be performed by a module (such as a chip) in the base station, or by a control subsystem that includes the base station functions. The control subsystem that includes the base station functions here may be a control center in the above-mentioned application scenarios such as smart grid, industrial control, smart transportation, and smart city. The functions of the terminal may also be performed by a module (such as a chip or modem) in the terminal, or by a device that includes the terminal functions.

[0060] Refer to FIG. 1 b , which is another schematic diagram of a wireless communication system applicable to an embodiment of the present application.

[0061] As shown in Figure 1b, the wireless communication system includes a RAN intelligent controller (RIC). As an example, RIC can be used to implement AI-related functions. As an example, the RIC includes a near-real time RIC (near-real time RIC, near-RT RIC) and a non-real time RIC (non-real time RIC, Non-RT RIC). Among them, the non-real-time RIC mainly processes non-real-time information, such as data that is not sensitive to delay, and the delay of the data can be in the order of seconds. Real-time RIC mainly processes near real-time information, such as data that is relatively sensitive to delay, and the delay of the data is in the order of tens of milliseconds.

[0062] The near real-time RIC is used for model training and reasoning. For example, it is used to train an AI model and use the AI ​​model for reasoning. The near real-time RIC can obtain network-side and / or terminal-side information from a RAN node (e.g., CU, CU-CP, CU-UP, DU, and / or RU) and / or a terminal. This information can be used as training data or reasoning data. Optionally, the near real-time RIC can deliver the reasoning result to the RAN node and / or the terminal. Optionally, the reasoning result can be exchanged between the CU and the DU, and / or between the DU and the RU. For example, the near real-time RIC delivers the reasoning result to the DU, and the DU sends it to the RU.

[0063] The non-real-time RIC is also used for model training and reasoning. For example, it is used to train an AI model and use the model for reasoning. The non-real-time RIC can obtain network-side and / or terminal-side information from RAN nodes (such as CU, CU-CP, CU-UP, DU and / or RU) and / or terminals. This information can be used as training data or reasoning data, and the reasoning results can be submitted to the RAN node and / or terminal. Optionally, the reasoning results can be exchanged between the CU and the DU, and / or between the DU and the RU. For example, the non-real-time RIC submits the reasoning results to the DU, and the DU sends it to the RU.

[0064] The near real-time RIC and non-real-time RIC may also be separately configured as a network element. Optionally, the near real-time RIC and non-real-time RIC may also be part of other devices. For example, the near real-time RIC is configured in a RAN node (e.g., a CU or DU), while the non-real-time RIC is configured in an OAM, a cloud server, a core network device, or other network device.

[0065] In practical applications, the wireless communication system may include multiple network devices (also called access network devices) and multiple terminal devices at the same time, without limitation. A network device may serve one or more terminal devices at the same time. A terminal device may also access one or more network devices at the same time. The embodiments of the present application do not limit the number of terminal devices and network devices included in the wireless communication system.

[0066] To facilitate understanding of the embodiments of the present application, the following describes several aspects of the embodiments of the present application:

[0067] (1) “Sending” and “receiving” in the embodiments of the present application indicate the direction of signal transmission. For example, “sending information to XX” can be understood as the destination of the information being XX, which can include direct sending through the air interface, as well as indirect sending through the air interface by other units or modules. “Receiving information from YY” can be understood as the source of the information being YY, which can include direct receiving from YY through the air interface, as well as indirect receiving from YY through the air interface from other units or modules. “Sending” can also be understood as the “output” of the chip interface, and “receiving” can also be understood as the “input” of the chip interface.

[0068] In other words, sending and receiving can be performed between devices, for example, between a network device and a terminal device, or can be performed within a device, for example, sending or receiving between components, modules, chips, software modules or hardware modules within the device through a bus, wiring or interface.

[0069] It is understandable that information may undergo necessary processing, such as encoding and modulation, between the source and destination of information transmission, but the destination can understand the valid information from the source. Similar expressions in this application can be understood similarly and will not be repeated here.

[0070] (2) In the embodiments of the present application, "indication" may include direct indication and indirect indication, and may also include explicit indication and implicit indication. The information indicated by a certain information (such as the indication information described below) is called information to be indicated. In the specific implementation process, there are many ways to indicate the information to be indicated, such as but not limited to, directly indicating the information to be indicated, such as the information to be indicated itself or the index of the information to be indicated. The information to be indicated may also be indirectly indicated by indicating other information, wherein the other information is associated with the information to be indicated; or only a part of the information to be indicated may be indicated, while the other part of the information to be indicated is known or agreed in advance. For example, the indication of specific information may be achieved by means of the arrangement order of each information agreed in advance (such as predefined by the protocol), thereby reducing the indication overhead to a certain extent. The present application does not limit the specific method of indication. It is understandable that for the sender of the indication information, the indication information can be used to indicate the information to be indicated, and for the receiver of the indication information, the indication information can be used to determine the information to be indicated.

[0071] In order to facilitate understanding of the relevant contents of the embodiments of the present application, some of the terms involved in the embodiments of the present application are explained below. This part is only for ease of understanding and cannot be regarded as a disclosure or specific limitation of the technical solution of the present application.

[0072] 1. Neural Networks

[0073] A neural network can be composed of neural units, which can be represented by x s The output of the operation unit can be shown in formula (1).

[0074] Where, s = 1, 2, ... n, n is a natural number greater than 1, W s is x s The weight of the neural unit, b is the bias of the neural unit. f is the activation function of the neural unit, which is used to introduce nonlinear characteristics into the neural network to convert the input signal in the neural unit into the output signal. The output signal of the activation function can be used as the input of the next convolutional layer. The activation function can be a sigmoid function. A neural network is a network formed by connecting many of the above-mentioned single neural units together, that is, the output of one neural unit can be the input of another neural unit. The input of each neural unit can be connected to the local receptive field of the previous layer to extract the features of the local receptive field. The local receptive field can be an area composed of several neural units.

[0075] It should be noted that the neural network model mentioned in this application may be one or more of a neural network model, a deep neural network (DNN) network model, a convolutional neural network (CNN) network model, a recurrent neural network (RNN) network model, a generative adversarial network network model, or a variation (or combination) of a combination thereof, and this application does not specifically limit this.

[0076] 2. Intelligent air interface technology

[0077] Intelligent air interface technology can be understood as a technology that applies a neural network model in a wireless communication system to optimize the performance of the wireless communication system. Please refer to Figure 2, which shows a wireless communication system based on a neural network transmitter and receiver (hereinafter collectively referred to as a transceiver) provided by this application. In the system shown in Figure 2, the transceiver that applies the neural network model can optimize the performance of transmitting and receiving signals through data-driven training.

[0078] It can be understood that optimizing the performance of a transceiver by applying a neural network model is to optimize the channel coding, modulation, waveform, pilot and other modules used for signal processing in the transceiver by using a neural network model. For example, please refer to FIG3 , which is a schematic diagram of optimizing different modules in a transceiver using several neural network models provided in an embodiment of the present application. Among them, FIG3a is a schematic diagram of optimizing a modulation module and a waveform by using a neural network model; FIG3b is a schematic diagram of optimizing a coding module and a modulation module by using a neural network model; and FIG3c is a schematic diagram of optimizing a reference signal by using a neural network model.

[0079] In 3a of Figure 3, after the transmitted bits are encoded by forward error correction (FEC) to obtain coded bits, the coded bits are modulated by the modulation neural network model (i.e., the NN-Mod module in 3a of Figure 3), mapped to the physical resources, transformed into time domain signals by inverse fast Fourier transform (IFFT), and then processed by the time signal neural network model (i.e., the T-NN module in 3a of Figure 3) before being sent; the received signal is first processed by the time signal neural network model, transformed into frequency domain signals by fast Fourier transform (FFT), and then demodulated by the demodulation neural network model (i.e., the NN-DeMod module in 3a of Figure 3) before being sent to FEC decoding.

[0080] In 3b of Figure 3, the bits to be transmitted are processed by the coding modulation neural network model (i.e., the NN-CoMo module in 3b of Figure 3) to obtain modulation symbols, which are then processed by modules such as IFFT and sent; the receiving device sends the signals processed by modules such as FFT directly to the demodulation and decoding neural network model (i.e., the NN-DeCoMo module in 3b of Figure 3) for processing to obtain estimated transmission bits.

[0081] In 3c of Figure 3, the transmitting end generates a reference signal based on the reference signal neural network model (i.e., the NN-RS module in 3c of Figure 3), and the receiving end processes the received signal based on the reference signal and the corresponding channel estimation neural network model (i.e., the NN-CE module in 3c of Figure 3) to obtain an estimated channel.

[0082] 3. Intelligent air interface training

[0083] The training of the intelligent air interface can be understood as the training of the neural network model used in the wireless communication system. Please refer to Figure 4, which is a schematic diagram of the training process of an intelligent air interface provided by this application. Among them, the sending neural network model performs forward reasoning on the training data, and sends the forward reasoning result to the receiving neural network model; the receiving neural network model performs forward reasoning and reverse gradient calculation based on the forward reasoning result to obtain a feedback result; the receiving neural network model sends the feedback result to the sending neural network model; the sending neural network model performs reverse gradient calculation based on the feedback result, and updates the model parameters of the sending neural network model based on the gradient obtained by the reverse gradient calculation. Exemplarily, the update of the tth round of the sending neural network model is as shown in formula (2).

[0084] Among them, w t+1 is the parameter of the t+1th round, that is, the parameter after the tth round update; w t is the parameter of the tth round; z k is the output of the sending neural network corresponding to the kth training sample, l k is the loss function corresponding to the tth round, K is the number of samples in this round of training; η t is the learning rate; is the gradient of the loss function with respect to the output of the neural network model.

[0085] That is, during any round of training, the neural network model is first sent for forward inference, obtaining the forward inference results and intermediate results. The intermediate inference results are cached. Then, the loss function is calculated from the last layer, and the gradient is calculated layer by layer according to the chain rule. The gradient calculation process requires the forward inference results of each layer. The gradients calculated for multiple training data are added together and used to send the model parameter updates to the neural network model.

[0086] Usually, the training process shown in FIG4 is to deploy the neural network model in the equipment of the wireless communication system after the training of the neural network model is completed in the training system. In order to improve the adaptability of the neural network model to the actual application scenario, it is proposed to deploy the neural network model on the transmitting device or the receiving device, and then update and adjust the model parameters of the neural network model in the actual communication scenario (i.e., the information transmission between the transmitting device and the receiving device is affected by the communication environment). However, during such a training process, the transmitting device or the receiving device may still be performing other services, resulting in the resources (including but not limited to computing resources, storage space, etc.) that the transmitting device or the receiving device can use for the neural network model training being not fixed, thereby resulting in the instability of the training time during the neural network model training process, affecting the performance of the wireless communication system to which the neural network model is applied.

[0087] In order to optimize the performance of a wireless communication system that uses a neural network model, the present application provides a neural network model training method and a communication device. The neural network model training method and communication device provided by the present application are further described below with reference to the accompanying drawings:

[0088] Please refer to Figure 5, which is a flow chart of a neural network model training method provided by an embodiment of the present application. As shown in Figure 5, the neural network model training method includes the following S501 to S504. The execution subject of the method shown in Figure 5 can be a sending end device and a receiving end device, or the execution subject of the method shown in Figure 5 can be a module in the sending end device and a module in the receiving end device, or the execution subject of the method shown in Figure 5 can be a chip of the sending end device and a chip of the receiving end device. Figure 5 is illustrated by taking the sending end device and the receiving end device as the execution subject of the method as an example. It should be noted that the sending end device (or receiving end device) mentioned in this application can be the network device shown in Figure 1a, or it can be the terminal device shown in Figure 1a, and this application does not make specific restrictions. Among them:

[0089] S501: A transmitting device obtains a target feedback mode and a target update mode.

[0090] S502: The receiving device obtains a target feedback mode.

[0091] It should be noted that when training the neural network model deployed on the sending end device (i.e., the first neural network model mentioned in this application), the sending end device and the receiving end device reach a consensus on the target feedback mode. It can be understood that when training the first neural network model, the target feedback mode obtained by the sending end device in S501 (or understood to be known) is the same as the target feedback mode obtained by the receiving end device in S502 (or understood to be known), and this application does not limit the execution order of S501 and S502. Optionally, when training the first neural network model, the sending end device and the receiving end device can also reach a consensus on the target update mode during the training process. It should be noted that, for ease of understanding, the following text takes the example of the sending end device and the receiving end device reaching a consensus on the target feedback mode and the target update mode during the training process as an example for explanation.

[0092] That is, after a device (hereinafter referred to as a determining device for ease of description) determines a target feedback mode and a target update mode, it indicates the target feedback mode and target update mode to a transmitting device and / or a receiving device, so that the transmitting device and the receiving device reach a consensus on the target feedback mode and target update mode. The determining device may be a transmitting device, a receiving device, or a device that has a communication connection with the transmitting device and the receiving device, and this application does not specifically limit this. It can be understood that when the determining device is neither a sending device nor a receiving device (for example, a device having a communication connection with the sending device and the receiving device), after the determining device determines the target feedback mode and the target update mode, it is necessary to indicate the target feedback mode and the target update mode to the sending device and the receiving device; when the determining device is the sending device, after the determining device determines the target feedback mode and the target update mode, it is only necessary to indicate the target feedback mode and / or target update mode to the receiving device; when the determining device is the receiving device, after the determining device determines the target feedback mode and the target update mode, it is only necessary to indicate the target feedback mode and / or target update mode to the sending device.

[0093] The following first describes how to determine the target feedback mode and the target update mode.

[0094] In one possible implementation, the determination device obtains one or more of the first capability information of the transmitting device, the second capability information of the receiving device, and / or the first channel information; further, the determination device determines a target feedback mode based on one or more of the first capability information, the second capability information, and / or the first channel information; and determines a target update mode based on one or more of the first capability information, the second capability information, and / or the first channel information.

[0095] The first capability information indicates one or more of the computing speed of the transmitting device, the size of the buffer space of the transmitting device, the forward reasoning time for the reference model of the transmitting device, and / or the reverse gradient calculation time. The second capability information indicates one or more of the computing speed of the receiving device, the size of the buffer space of the receiving device, the forward reasoning time for the reference model of the receiving device, and / or the reverse gradient calculation time. The first channel information is information corresponding to the channel between the transmitting device and the receiving device, and the first channel information indicates one or more of the wireless resources, channel quality, communication rate, and / or communication delay of the channel.

[0096] It should be noted that the reference model mentioned in this application is a model for measuring the computing time (such as forward reasoning time and / or reverse gradient calculation time) of a device (including a transmitting device and a receiving device). This application does not specifically limit the name of the reference model, nor does it specifically limit the specific model structure of the reference model. It should also be noted that the determination device for determining the target feedback mode and the determination device for determining the target update mode can be the same or different, and this application does not make specific limitations on this; the following text is only for ease of understanding, and an exemplary explanation is given by taking the example that the determination device for determining the target feedback mode and the determination device for determining the target update mode are the same. It should also be noted that the information used by the determination device to determine the target feedback mode and the information used to determine the target update mode can be the same or different, and this application does not make specific limitations on this; for example, the determination device determines the target feedback mode based on the second capability information and the first channel information, and determines the target update mode based on the first capability information.

[0097] For ease of understanding, this application takes the aforementioned determination device as the network device N determining the target feedback mode and the target update mode based on the first capability information, the second capability information and the first channel information as an example, and provides the following three example situations for illustrative explanation, which should not be regarded as a specific limitation of this application, that is, the determination device can also be a terminal device.

[0098] In Example 1, network device N is a transmitting device, and the receiving device is terminal device A. In this case, network device N obtains the first channel information and the second capability information sent by terminal device A, and determines the target feedback mode and the target update mode based on the first capability information, the second capability information and the first channel information of network device N. Furthermore, network device N sends indication information indicating the target feedback mode and the target update mode to terminal device A. It should be noted that the way in which network device N obtains the first channel information can be that network device N sends a reference signal to the terminal device, and the terminal device performs channel measurement based on the reference signal to obtain the first channel information. Further, the terminal device sends the first channel information to network device N; unless otherwise specified below, the first channel information can be obtained in this way.

[0099] In Example 2, network device N is the receiving device, and the sending device is terminal device B. In this case, network device N obtains the first channel information and the first capability information sent by terminal device B, and determines the target feedback mode and the target update mode based on the second capability information, the first capability information, and the first channel information of network device N. Furthermore, network device N sends indication information to terminal device B indicating the target feedback mode and the target update mode.

[0100] In Example 3, the transmitting device is terminal device B, the receiving device is terminal device A, and network device N is a network device that provides services to terminal device B and terminal device A. In this case, network device N receives the first capability information sent by terminal device B and the second capability information sent by terminal device A, and obtains first channel information of the first channel measured by terminal device A or terminal device B (i.e., the channel between terminal device A and terminal device B). Furthermore, network device N determines a target feedback mode and a target update mode based on the first capability information, the second capability information, and the first channel information, and sends indication information indicating the target feedback mode and target update mode to terminal device A and terminal device B.

[0101] The target feedback mode and target update mode are described in detail below.

[0102] 1. The target feedback mode includes the first feedback mode or the second feedback mode.

[0103] In the first feedback mode, after receiving the forward reasoning result from the sending device, the receiving device will send the feedback result of the forward reasoning result to the sending device after a fixed time interval, that is, the time interval between the reception time of the forward reasoning result and the sending time of the feedback result is fixed (or it can be understood that the time interval is static).

[0104] For example, as shown in FIG6 , when the transmitting device sends a forward reasoning result to the receiving device in time slot n, the receiving device sends a feedback result for the forward reasoning result to the transmitting device in time slot n+G, and the time interval between time slot n+G and time slot n is G time slots; when the transmitting device sends a forward reasoning result to the receiving device in time slot m, the receiving device sends a feedback result for the forward reasoning result to the transmitting device in time slot m+G, and the time interval between time slot m+G and time slot m is G time slots. It should be noted that this application uses time slots as the unit of time interval for illustrative purposes and should not be regarded as a specific limitation of this application; that is, the unit of time interval can also be a symbol, a symbol group, a frame, or a subframe, etc.; the same applies to the entire text.

[0105] It can be understood that when at least one of the following situations exists, the target feedback mode can be determined to be the first feedback mode: Situation 1, when the first capability information indicates that the available resources (including but not limited to computing resources, storage space, etc.) used by the sending device for neural network model training are fixed (or understood as having sufficient available resources), the time taken by the sending device to calculate the forward reasoning result each time is almost the same (or understood as not changing significantly); Situation 2, when the second capability information indicates that the available resources (including but not limited to computing resources, storage space, etc.) used by the receiving device for neural network model training are fixed (or understood as having sufficient available resources), the time taken by the receiving device to obtain the feedback result based on the forward reasoning result is almost the same (or understood as not changing significantly); Situation 3, when the first channel information indicates that the current channel environment is reliable (for example, low communication delay, high signal-to-noise ratio, etc.), the time taken to transmit the feedback result from the receiving device to the sending device is almost the same (or understood as not changing significantly).

[0106] In one possible implementation, when the target feedback mode is the first feedback mode, the transmitting device may also determine the feedback time based on the time interval and the sending time of the forward reasoning result; wherein the time interval is the interval between the sending time of the forward reasoning result and the feedback time of the feedback result (for example, G time slots in the example corresponding to Figure 6), and the time interval is determined based on one or more of the first capability information, the second capability information, or the first channel information.

[0107] That is to say, when the determination device determines that the target feedback mode is the first feedback mode, it can also determine the time interval corresponding to the target feedback mode based on at least one of the first capability information, the second capability information or the first channel information. Further, the determination device indicates to the transmitting device and / or the receiving device that the target feedback mode is the first feedback mode, and indicates the time interval corresponding to the first feedback mode. The subsequent transmitting device can determine the feedback time of the feedback result based on the time interval and the sending time of the forward reasoning result, and the receiving device can also determine the feedback time of the feedback result based on the sending time of the forward reasoning result (or understand it as the sending time). It should be noted that, for ease of understanding, the following text of this application uses the example of ignoring the transmission time of information (including forward reasoning results and feedback results) for explanation, which cannot be regarded as a specific limitation of this application.

[0108] In one possible implementation, when the target feedback mode is the first feedback mode, the transmitting device may also periodically send forward reasoning results to the receiving device according to a first period. The first period may be determined by the aforementioned determination device based on one or more of the first capability information, the second capability information, and / or the first channel information. It can be understood that when at least one of the aforementioned situations 1 to 3 exists, the transmitting device may periodically send forward reasoning results to the receiving device according to the first period.

[0109] In the second feedback mode, after receiving the forward reasoning result from the sending device, the receiving device will not send the feedback result of the forward reasoning result to the sending device after a fixed time interval, that is, the time interval between the reception time of the forward reasoning result and the sending time of the feedback result is not fixed (or it can be understood that the time interval is dynamic).

[0110] For example, as shown in Figure 7, when the transmitting device sends a forward reasoning result to the receiving device at time slot n, the receiving device sends feedback regarding the forward reasoning result to the transmitting device at time slot n+K, with the time interval between time slot n+K and time slot n being K time slots. When the transmitting device sends a forward reasoning result to the receiving device at time slot m, the receiving device sends feedback regarding the forward reasoning result to the transmitting device at time slot m+L, with the time interval between time slot m+L and time slot m being L time slots. Typically, K and L are unequal positive integers, but in certain circumstances (which can be understood as low-probability events or occasional events), K and L can also be equal.

[0111] In a possible implementation, when the target feedback mode is the second feedback mode, the transmitting end device receives first indication information from the receiving end device, where the first indication information is used to indicate a feedback time of a feedback result.

[0112] For example, taking the receiving device as network device N and the sending device as terminal device A, after terminal device A sends a forward reasoning result to network device N, terminal device A cannot independently determine when network device N will provide feedback regarding the forward reasoning result. In this case, network device N can estimate the feedback time based on its current capability information; or, after network device N obtains feedback based on the forward reasoning result, it can send first indication information to terminal device A, indicating the feedback time. Subsequently, network device N will send feedback to terminal device A at the feedback time.

[0113] In one possible implementation, when the target feedback mode is the second feedback mode, the transmitting device receives a request message from the receiving device, where the request message is used to request sending a feedback result; and sends a confirmation message to the receiving device, where the confirmation message includes first indication information, where the first indication information is used to indicate a feedback time.

[0114] For example, in the case of a receiving device, terminal device B, and a sending device, network device N, after network device N sends a forward reasoning result to terminal device B, network device N cannot independently determine when terminal device B will provide feedback regarding that forward reasoning result. In this case, terminal device B can estimate the time it will provide feedback based on its current capabilities. Alternatively, after receiving feedback based on the forward reasoning result, terminal device B can send a request message to network device N requesting the feedback result. Furthermore, network device N can send a confirmation message to terminal device B, informing terminal device B of the feedback time.

[0115] In one possible implementation, when the target feedback mode is the second feedback mode, before the transmitting device transmits the forward reasoning result to the receiving device, the transmitting device first transmits the forward reasoning result's sending time to the receiving device, and then transmits the forward reasoning result to the receiving device at the sending time. For example, the transmitting device may estimate the sending time of the forward reasoning result based on its current capability information; or, after obtaining the forward reasoning result, the transmitting device transmits the forward reasoning result's sending time to the receiving device, and then transmits the forward reasoning result to the receiving device at the sending time.

[0116] 2. The target update mode includes the first update mode or the second update mode.

[0117] In the first update mode, the transmitting end device updates the model parameters of the first neural network model after receiving N feedback results, where N is an integer greater than or equal to 1. In other words, the first update mode indicates updating after receiving N feedback results.

[0118] In the second update mode, the transmitting end device updates the model parameters of the first neural network model after receiving the update instruction information. In other words, the second update mode instructs the transmitting end device to update the model parameters of the first neural network model after receiving the third instruction information (used to instruct to update the model parameters of the first neural network model).

[0119] S503. At a feedback time corresponding to the target feedback mode, the transmitting device receives a feedback result for a forward reasoning result from the receiving device, where the forward reasoning result is obtained based on a first neural network model deployed in the transmitting device.

[0120] That is, at the feedback time corresponding to the target feedback mode, the receiving end device sends the feedback result to the transmitting end device. Also, at the feedback time corresponding to the target feedback mode, the transmitting end device receives the feedback result from the receiving end device.

[0121] It should be understood that in this application, the neural network model deployed in the sending device is referred to as the first neural network model. The first neural network model performs forward reasoning on the training data based on the first model parameters to obtain a forward reasoning result. The sending device sends the forward reasoning result to the receiving device. After receiving the forward reasoning result, the receiving device calculates a feedback result based on the forward reasoning result, and sends the feedback result to the sending device at the feedback time corresponding to the target feedback mode. Among them, the first model parameter can be understood as the model parameter of the first neural network model before the first neural network model is updated when S503 to S504 are executed to update the first neural network model. Among them, the training data can be understood as data used for training the neural network model; for example, the training data can be training data with a label (or called an annotation), which is used to update the model parameters of the neural network model (including the first neural network model or the second neural network model mentioned in this application). When the training data has a label, the label is related to the task of the neural network model, and the label can be understood as the result obtained after the expected neural network model processes the training data. For example, if the neural network model is used for data recovery (i.e., the task of the neural network model is data recovery), the label of the training data is the true value of the training data (or the data to be recovered); if the neural network model is used for data classification (i.e., the task of the neural network model is data classification), the label of the training data is the true category of the training data (or the data to be classified). This application does not specifically limit the specific form of the training data.

[0122] In one example, a transmitting device and a receiving device determine that the target feedback mode is a first feedback mode, and the time interval corresponding to the first feedback mode is G time slots. In this case, as shown in FIG6 , after the transmitting device obtains the forward reasoning result, it sends the forward reasoning result to the receiving device in time slot n. Furthermore, based on the time interval (i.e., G time slots), the transmitting device determines that the time when the receiving device sends the forward reasoning result is time slot n+G. The transmitting device receives feedback on the forward reasoning result from the receiving device in time slot n+G.

[0123] In another example, the transmitting device and the receiving device determine that the target feedback mode is the second feedback mode. In this case, as shown in FIG7 , after the transmitting device obtains the forward reasoning result, it sends the forward reasoning result to the receiving device in time slot n. The receiving device receives the forward reasoning result. After the receiving device obtains the feedback result based on the forward reasoning result, or after the receiving device determines the time to obtain the feedback result based on its own capability information (which can be understood as the feedback time), the receiving device sends a first indication information to the transmitting device. The first indication information indicates that the feedback time of the feedback result is time slot n+K. Furthermore, the transmitting device receives the feedback result for the forward reasoning result from the receiving device in time slot n+K.

[0124] In one possible implementation, a transmitting device sends multiple forward reasoning results to a receiving device, and the receiving device receives multiple feedback results. That is, when there are multiple forward reasoning results and multiple feedback results, the forward reasoning results and the feedback results correspond one-to-one. In this case, the transmitting device may also receive second indication information from the receiving device, indicating the corresponding relationship between each feedback result and the forward reasoning result.

[0125] For example, when the target feedback mode is the second feedback mode, the transmitting device sends the first forward reasoning result to the receiving device at moment 1, and sends the second forward reasoning result to the receiving device at moment 2. The receiving device obtains the first feedback result based on the first forward reasoning result, and obtains the second feedback result based on the second forward reasoning result. If moment 1 is before moment 2, the time when the receiving device sends the first feedback result (i.e., moment 3 mentioned in this application) is after moment 2, in this case, when the receiving device sends the first feedback result to the transmitting device, it indicates that the first feedback result is the feedback result corresponding to the first forward reasoning result; when the second feedback result is sent to the transmitting device at moment 4 (i.e., a moment after moment 2), it indicates that the second feedback result is the feedback result corresponding to the second forward reasoning result.

[0126] In one possible implementation, the first neural network model corresponds to at least one training process, and each training process corresponds to at least one forward reasoning result. In this case, the receiving device may indicate the correspondence between the feedback result and the forward reasoning result by indicating the identifier of the training process corresponding to the feedback result.

[0127] Exemplarily, the training process corresponding to the first neural network device includes training process 1 and training process 2, wherein training process 1 includes training data 1 and training data 2, and training process 2 includes training data 3 and training data 4. Training data 1 to training data 4 are input into the first neural network model of the sending end device to obtain forward reasoning results 1 to forward reasoning results 4, respectively. At time 1, the sending end device sends forward reasoning result 1 and forward reasoning result 2 to the receiving end device; at time 2 (later than time 1), the sending end device sends forward reasoning result 3 and forward reasoning result 4 to the receiving end device. Furthermore, the receiving end device obtains feedback results 1 to feedback results 4, respectively, based on forward reasoning results 1 to forward reasoning results 4. The corresponding relationship between each training data, forward reasoning result, and feedback result is shown in Table 1.

[0128] Table 1

[0129] In this case, when the receiving device sends feedback results 1 and 2 to the sending device at time 3 (later than time 2), it indicates that the identifier of the training process corresponding to feedback results 1 and 2 is training process 1. When the receiving device sends feedback results 3 and 4 to the sending device at time 4 (later than time 2), it indicates that the identifier of the training process corresponding to feedback results 3 and 4 is training process 2. Furthermore, after receiving the feedback results, the sending device can know the corresponding relationship between the feedback results and the forward reasoning results based on the identifier of the training process.

[0130] It should be noted that the training process referred to in this application includes: the process of the first neural network model performing forward inference on a set of training data (i.e., the training data of the training process), the process of receiving feedback results corresponding to the set of training data, and the process of calculating gradients based on the forward inference results, intermediate results, and feedback results of the set of training data. This application does not specifically limit the name of the training process.

[0131] S504: The transmitting device updates the model parameters of the first neural network model based on the target update mode and the feedback result.

[0132] After receiving the feedback result, the transmitting device calculates the gradient corresponding to the forward reasoning result based on the feedback result, the forward reasoning result corresponding to the feedback result, and the intermediate result corresponding to the forward reasoning result, and caches the gradient corresponding to the forward reasoning result. Furthermore, the transmitting device updates the model parameters of the first neural network model based on the gradient based on the update timing (or update condition) indicated by the target update mode.

[0133] It should be noted that the forward reasoning result mentioned in this application is the result to be sent to the receiving device after calculation (or understanding as reasoning) by the first neural network model; or it can be understood as the calculation result obtained by the last layer of neural network in the first neural network model in the process of calculating the forward reasoning result. The forward reasoning result mentioned in this application can also be called a calculation result, or an inference result, etc. in other schemes, and this application does not specifically limit its name. The intermediate result mentioned in this application is the information generated by the first neural network model in the process of calculating the forward reasoning result, or it can be called an intermediate state; or it can also be understood as the calculation result obtained by the other layers of neural network in the first neural network model except the last layer of neural network in the process of calculating the forward reasoning result. The feedback result mentioned in this application can be the loss result of the inference calculation of the neural network model of the receiving device (that is, the second neural network model mentioned in this application), or the gradient obtained by the reverse gradient calculation of the second neural network model. This application does not specifically limit this.

[0134] In one example, the target update mode is the aforementioned first update mode; in this first update mode, the transmitting device updates the model parameters of the first neural network model after receiving one feedback result. It can be understood that the update timing indicated by the target update mode is: every time one feedback result is received, the model parameters of the first neural network model are updated. In this case, as shown in Figure 8, the transmitting device sends the first forward reasoning result (obtained based on the model parameter W) to the receiving device at time slot n, and the intermediate result corresponding to the first forward reasoning result is the first intermediate result (obtained based on the model parameter W). After the transmitting device receives the first feedback result obtained for the first forward reasoning result from the receiving device at time slot n+G, it obtains the first gradient based on the first feedback result, the first forward reasoning result and the first intermediate result; and based on the first gradient, updates the model parameter W of the first neural network model to the model parameter W1. The transmitting device sends the second forward reasoning result (obtained based on the model parameter W1) to the receiving device in time slot m, and the intermediate result corresponding to the second forward reasoning result is the second intermediate result (obtained based on the model parameter W1); after the transmitting device receives the second feedback result obtained for the second forward reasoning result from the receiving device in time slot m+G, it obtains a second gradient based on the second feedback result, the second forward reasoning result and the second intermediate result; and based on the second gradient, updates the model parameter W1 of the first neural network model to the model parameter W2.

[0135] In another example, the target update mode is the aforementioned second update mode. It can be understood that the update timing indicated by the target update mode is: upon receiving the update indication information (i.e., the third indication information mentioned in the application) from the receiving end device, the model parameters of the first neural network model are updated. In this case, as shown in FIG9 , the transmitting end device sends the first forward reasoning result (obtained based on the model parameter W) to the receiving end device at time slot n, and the intermediate result corresponding to the first forward reasoning result is the first intermediate result (obtained based on the model parameter W); the transmitting end device receives the first feedback result obtained for the first forward reasoning result from the receiving end device at time slot n+G; the transmitting end device calculates the first gradient based on the first feedback result, the first forward reasoning result, and the first intermediate result, and caches the first step. Then, the transmitting device sends a second forward reasoning result (obtained based on model parameter W) to the receiving device at time slot m. The intermediate result corresponding to the second forward reasoning result is the second intermediate result (obtained based on model parameter W). After the transmitting device receives a second feedback result obtained from the receiving device for the second forward reasoning result at time slot m+G, it calculates a second gradient based on the second feedback result, the second forward reasoning result, and the second intermediate result, and caches the second gradient. Furthermore, after the transmitting device receives the third indication information sent by the receiving device, the transmitting device updates the model parameter W of the first neural network model to model parameter W3 based on the first gradient and the second gradient.

[0136] In another example, the target update mode is the aforementioned second update mode. It can be understood that the update timing indicated by the target update mode is: upon receiving the update indication information (i.e., the third indication information mentioned in the application) from the receiving device, the model parameters of the first neural network model are updated. In this case, as shown in FIG10 , the transmitting device sends the first forward reasoning result (obtained based on the model parameter W) to the receiving device at time slot n, and the intermediate result corresponding to the first forward reasoning result is the first intermediate result (obtained based on the model parameter W). The transmitting device receives indication information 1 from the receiving device, and the indication information 1 indicates that the feedback time is time slot n+K. Further, the transmitting device receives the first feedback result obtained from the receiving device for the first forward reasoning result at time slot n+K; the transmitting device calculates the first gradient based on the first feedback result, the first forward reasoning result, and the first intermediate result, and caches the first gradient. Then, the transmitting device sends the second forward reasoning result (obtained based on the model parameter W) to the receiving device at time slot m, and the intermediate result corresponding to the second forward reasoning result is the second intermediate result (obtained based on the model parameter W). The transmitting device receives indication information 2 from the receiving device, indicating that the feedback time is time slot m+L. Furthermore, after receiving a second feedback result obtained from the receiving device for the second forward reasoning result at time slot m+L, the transmitting device calculates a second gradient based on the second feedback result, the second forward reasoning result, and the second intermediate result, and caches the second gradient. Furthermore, after receiving a third indication information from the receiving device, the transmitting device updates the model parameter W of the first neural network model to model parameter W4 based on the first gradient and the second gradient.

[0137] It should be noted that the present application does not limit the timing of the receiving device sending the third indication information. The receiving device may send the third indication information after sending N feedback results. Alternatively, after the receiving device receives a forward reasoning result X with a different data distribution (or data characteristics, or data source) than the previous Q (where Q is a positive integer) forward reasoning results, and after sending feedback results of the forward reasoning result X, the receiving device may send the third indication information to instruct the sending device to update the model parameters of the first neural network model, which is beneficial for avoiding model overfitting.

[0138] It can be seen that by implementing the neural network model training method described in Figure 5 of the present application, after the first neural network model is deployed on the transmitting device, before the transmitting device trains the first neural network model, the transmitting device can first know the target feedback mode of the receiving device sending the feedback result, and know the update mode of the neural network model deployed on the transmitting device, which is beneficial for the transmitting device to determine the training duration of the neural network model, and thus is beneficial for optimizing the performance of the wireless communication system that applies the neural network model.

[0139] It should be noted that the neural network model training method shown in FIG5 only describes the update process for the neural network model deployed on the transmitting device (i.e., the aforementioned first neural network model), and does not limit the update process for the neural network model deployed on the receiving device (i.e., the second neural network model mentioned in this application). In other words, during the update process of the first neural network model, the receiving device can update the model parameters of the second neural network model based on the forward reasoning results from the transmitting device; the receiving device may also not update the model parameters of the second neural network model; this application does not specifically limit this.

[0140] For ease of understanding, this application illustrates the example of a receiving device updating the model parameters of a second neural network model. In one possible implementation, the receiving device obtains a target update mode corresponding to the receiving device. To distinguish the target update modes in FIG. 5 , the target update mode corresponding to the receiving device will be replaced by the target update mode Rx, and the target update mode corresponding to the transmitting device will be replaced by the target update mode Tx. The receiving device then updates the model parameters of the second neural network model based on the target update mode Rx.

[0141] Among them, the target update mode Rx includes a third update mode and a fourth update mode. When the target update mode Rx is the third update mode, the receiving end device updates the model parameters of the second neural network model after receiving M forward reasoning results. When the target update mode Rx is the fourth update mode, the receiving end device updates the model parameters of the second neural network model after obtaining (including generating or receiving) the fourth indication information (used to indicate the update of the model parameters of the second neural network model). Among them, M is the same as the aforementioned N or different, and this application does not limit this; the fourth indication information and the aforementioned third indication information can be the same or different, and this application does not limit this either.

[0142] It should be noted that the manner in which the receiving device acquires the target update mode Rx can be referenced to the manner in which the transmitting device acquires the target update mode in S501 above, and will not be further described here. That is, the receiving device may determine the target update mode Rx based on at least one of the first capability information, the second capability information, and / or the first channel information. Alternatively, a device (e.g., the transmitting device, a network device providing services to the receiving device, etc.) communicating with the receiving device may determine the target update mode Rx and then indicate the target update mode Rx to the receiving device.

[0143] Below, the present application provides several complete examples of updating the first neural network model according to the target update mode Tx and updating the second neural network model according to the target update mode Rx to help understand the solution provided by the present application, which should not be regarded as a specific limitation of the present application.

[0144] For example, the target update mode Tx is the first update mode, the target update mode Rx is the third update mode, the model parameter of the first neural network model before updating is the model parameter W, and the model parameter of the second neural network model before updating is the model parameter W'. Continuing with the example corresponding to Figure 8 above, M and N are both 1. In this case, the transmitting device sends the first forward reasoning result to the receiving device at time slot n, and the intermediate result corresponding to the first forward reasoning result is the first intermediate result (obtained based on the model parameter W). The receiving device calculates the first feedback result of the first forward reasoning result through the second neural network model corresponding to the model parameter W'. The transmitting device receives the first feedback result from the receiving device at time slot n+G, and updates the model parameter W of the first neural network model to the model parameter W1 based on the first feedback result, the first forward reasoning result and the first intermediate result. In addition, the receiving device updates the model parameter W' of the second neural network model to the model parameter W1' based on the first forward reasoning result. Subsequently, the transmitting device sends the second forward reasoning result (obtained based on the model parameter W1) to the receiving device at time slot m, and the intermediate result corresponding to the second forward reasoning result is the second intermediate result. The receiving device calculates the second feedback result of the second forward reasoning result through the second neural network model corresponding to the model parameter W1'. After the transmitting device receives the second feedback result from the receiving device at time slot m+G, it updates the model parameter W1 of the first neural network model to the model parameter W2 based on the second feedback result, the second forward reasoning result and the second intermediate result. In addition, the receiving device updates the model parameter W1' of the second neural network model to the model parameter W2' based on the second forward reasoning result.

[0145] For another example, the target update mode Tx is the second update mode, the target update mode Rx is the fourth update mode, the model parameter of the first neural network model before updating is the model parameter W, and the model parameter of the second neural network model before updating is the model parameter W'. Continuing with the example corresponding to Figure 9 above. The transmitting device sends a first forward reasoning result to the receiving device at time slot n, and the intermediate result corresponding to the first forward reasoning result is the first intermediate result (obtained based on the model parameter W). The receiving device calculates the first feedback result of the first forward reasoning result through the second neural network model corresponding to the model parameter W'. The transmitting device receives the first feedback result from the receiving device at time slot n+K, and calculates the first gradient based on the first feedback result, the first forward reasoning result and the first intermediate result, and caches the first gradient. The receiving device stores the first forward reasoning result. Then, the transmitting device sends the second forward reasoning result (obtained based on the model parameter W) to the receiving device at time slot m, and the intermediate result corresponding to the second forward reasoning result is the second intermediate result; after the transmitting device receives the second feedback result obtained for the second forward reasoning result from the receiving device at time slot m+L, it calculates the second gradient based on the second feedback result, the second forward reasoning result and the second intermediate result, and caches the second gradient. The receiving device stores the second forward reasoning result. After the transmitting device receives the third indication information sent by the receiving device, the transmitting device updates the model parameter W of the first neural network model to the model parameter W3 based on the first gradient and the second gradient. After the receiving device obtains the fourth indication information, the receiving device updates the model parameter W' of the second neural network model to the model parameter W3' based on the first forward reasoning result and the second forward reasoning result.

[0146] Please refer to Figure 11, which shows a structural diagram of a communication device 1100 of an embodiment of the present application. The communication device shown in Figure 11 can be a sending end device, or a device in a sending end device, or a device that can be used in combination with a sending end device. The communication device shown in Figure 11 may include a communication unit 1101 and a processing unit 1102; the communication device shown in Figure 11 may be a receiving end device, or a device in a receiving end device, or a device that can be used in combination with a receiving end device. The communication device shown in Figure 11 may include a communication unit 1101 and a processing unit 1102. Specifically, the processing unit 1102 is used to process data, and the data may be data received by the communication unit 1101, and the processed data may also be sent by the communication unit 1101;

[0147] In one embodiment, the communication device 1100 is a transmitting end device, a device in a transmitting end device, or a device that can be used in conjunction with the transmitting end device, wherein:

[0148] Communication unit 1101 is used to obtain a target feedback mode and a target update mode; at a feedback time corresponding to the target feedback mode, communication unit 1101 is also used to receive feedback results for forward reasoning results from a receiving device, where the forward reasoning results are obtained based on a first neural network model deployed in a sending device; processing unit 1102 is used to update model parameters of the first neural network model based on the target update mode and the feedback results.

[0149] In one possible implementation, the communication unit 1101 is used to obtain one or more of the first capability information of the transmitting device, the second capability information of the receiving device, or the first channel information, where the first channel information is information corresponding to the channel between the transmitting device and the receiving device; the processing unit 1102 is also used to determine the target feedback mode based on one or more of the first capability information, the second capability information, or the first channel information; the processing unit 1102 is also used to determine the target update mode based on one or more of the first capability information, the second capability information, or the first channel information.

[0150] In one possible implementation, the target feedback mode is the first feedback mode, and the processing unit 1102 is further used to determine the feedback time based on the time interval and the sending time of the forward reasoning result; wherein the time interval is the interval between the sending time of the forward reasoning result and the feedback time of the feedback result, and the time interval is determined based on one or more of the first capability information, the second capability information or the first channel information.

[0151] In a possible implementation, the target feedback mode is the second feedback mode, and the communication unit 1101 is further configured to receive first indication information from a receiving-end device, where the first indication information is used to indicate a feedback time.

[0152] In one possible implementation, the target feedback mode is the second feedback mode, and the communication unit 1101 is further used to receive a request message from the receiving device, where the request message is used to request sending a feedback result; the communication unit 1101 is further used to send a confirmation message to the receiving device, where the confirmation message includes first indication information, where the first indication information is used to indicate a feedback time.

[0153] In one possible implementation, the number of forward reasoning results is multiple, the number of feedback results is multiple, the forward reasoning results and the feedback results correspond one to one, and the communication unit 1101 is also used to receive second indication information from the receiving end device, and the second indication information is used to indicate the correspondence between each feedback result and each forward reasoning result.

[0154] In one possible implementation, the target update mode is a first update mode or a second update mode; wherein the first update mode indicates updating after receiving N feedback results, where N is an integer greater than or equal to 1; and the second update mode indicates updating after receiving third indication information, where the third indication information is used to indicate updating of model parameters of the first neural network model.

[0155] In one embodiment, the communication device 1100 is a receiving device, a device in a receiving device, or a device that can be used in combination with a receiving device, wherein: the communication unit 1101 is used to obtain a target feedback mode; at a feedback time corresponding to the target feedback mode, the communication unit 1101 is also used to send a feedback result for a forward reasoning result to a sending device; the forward reasoning result is obtained based on a first neural network model deployed in the sending device.

[0156] In one possible implementation, the communication unit 1101 is further used to obtain one or more of the first capability information of the transmitting device, the second capability information of the receiving device, or the first channel information, where the first channel information is information corresponding to the channel between the transmitting device and the receiving device; the processing unit 1102 is further used to determine the target feedback mode based on one or more of the first capability information, the second capability information, or the first channel information.

[0157] In one possible implementation, the target feedback mode is the first feedback mode, and the processing unit 1102 is further used to determine the feedback time based on the time interval and the sending time of the forward reasoning result; wherein the time interval is the interval between the sending time of the forward reasoning result and the feedback time of the feedback result, and the time interval is determined based on one or more of the first capability information, the second capability information or the first channel information.

[0158] In a possible implementation, the target feedback mode is the second feedback mode, and the communication unit 1101 is further configured to send first indication information to the transmitting end device, where the first indication information is used to indicate a feedback time.

[0159] In one possible implementation, the target feedback mode is the second feedback mode, and the communication unit 1101 is further used to send a request message to the sending device, where the request message is used to request sending a feedback result; the communication unit 1101 is further used to receive a confirmation message from the sending device, where the confirmation message includes first indication information, where the first indication information is used to indicate a feedback time.

[0160] In one possible implementation, the number of forward reasoning results is multiple, the number of feedback results is multiple, the forward reasoning results and the feedback results correspond one to one, and the communication unit 1101 is also used to send second indication information to the sending end device, and the second indication information is used to indicate the correspondence between each feedback result and each forward reasoning result.

[0161] In one possible implementation, the communication unit 1101 is also used to obtain a target update mode, which is a first update mode or a second update mode; wherein the first update mode indicates updating after receiving N feedback results, where N is an integer greater than or equal to 1; the second update mode indicates updating after receiving third indication information, where the third indication information is used to indicate updating the model parameters of the first neural network model.

[0162] In a possible implementation, the target update mode is the second update mode, and the communication unit 1101 is further configured to send third indication information to the transmitting end device.

[0163] In a possible implementation, the receiving device is deployed with a second neural network model, and the processing unit 1102 is further used to update the model parameters of the second neural network model according to the forward reasoning result.

[0164] For a more detailed description of the communication unit 1101 and the processing unit 1102 , reference may be made to the relevant description of the transmitting end device or the receiving end device in the method embodiment shown in FIG5 .

[0165] As shown in Figure 12, communication device 1200 includes a processor 1210 and an interface circuit 1220. Processor 1210 and interface circuit 1220 are coupled to each other. It is understood that interface circuit 1220 may be a transceiver or an input / output interface. Optionally, communication device 1200 may further include a memory 1230 for storing instructions executed by processor 1210, input data required by processor 1210 to execute instructions, or data generated by processor 1210 after executing instructions.

[0166] When the communication device 1200 is used to implement the method shown in FIG5 , the processor 1210 is used to implement the functions of the processing unit 1102 , and the interface circuit 1220 is used to implement the functions of the communication unit 1101 .

[0167] When the above-mentioned communication device is a chip applied to a transmitting device, the terminal chip implements the functions of the transmitting device in the above-mentioned method embodiment. The transmitting device chip receives information from the receiving device, which can be understood as the information being first received by other modules in the transmitting device (such as a radio frequency module or antenna) and then sent to the transmitting device chip by these modules. The transmitting device chip sends information to the receiving device, which can be understood as the information being first sent to other modules in the transmitting device (such as a radio frequency module or antenna) and then sent to the receiving device by these modules.

[0168] When the communication device is a chip used in a receiving device, the receiving device chip implements the functions of the receiving device in the above-mentioned method embodiment. The receiving device chip receives information from the transmitting device, which can be understood as the information being first received by other modules in the receiving device (such as a radio frequency module or antenna) and then transmitted to the receiving device chip by these modules. The receiving device chip sends information to the transmitting device, which can be understood as the information being transmitted to other modules in the receiving device (such as a radio frequency module or antenna) and then transmitted to the transmitting device by these modules.

[0169] In this application, when entity A sends information to entity B, it can be done directly from A to B or indirectly through another entity. Similarly, when entity B receives information from entity A, it can be done directly from entity B or indirectly through another entity. Entities A and B herein can be RAN nodes or terminals, or modules within a RAN node or terminal. The sending and receiving of information can be information exchange between a RAN node and a terminal, for example, between a base station and a terminal; the sending and receiving of information can also be information exchange between two RAN nodes, for example, between a CU and a DU; the sending and receiving of information can also be information exchange between different modules within a device, for example, between a terminal chip and other modules in the terminal, or between a base station chip and other modules within the base station.

[0170] It is understood that the processor in the embodiments of the present application may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0171] The method steps in the embodiments of the present application can be implemented in hardware or in software instructions that can be executed by a processor. The software instructions can be composed of corresponding software modules, and the software modules can be stored in random access memory, flash memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, registers, hard disk, mobile hard disk, CD-ROM or any other form of storage medium well known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. The storage medium can also be an integral part of the processor. The processor and storage medium can be located in an ASIC. In addition, the ASIC can be located in a base station or a terminal. The processor and storage medium can also exist in a base station or a terminal as discrete components.

[0172] In the above embodiments, all or part of the embodiments may be implemented using software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are performed in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user device, or other programmable device. The computer program or instructions may be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, the computer program or instructions may be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; an optical medium, such as a digital video disk; or a semiconductor medium, such as a solid-state drive. The computer-readable storage medium may be a volatile or nonvolatile storage medium, or may include both volatile and nonvolatile types of storage media.

[0173] In the various embodiments of the present application, unless otherwise specified or there is a logical conflict, the terms and / or descriptions between different embodiments are consistent and can be referenced by each other. The technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationships.

[0174] In this application, "at least one" means one or more, and "more" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. In the text description of this application, the character " / " generally indicates that the previous and next associated objects are in an "or" relationship; in the formula of this application, the character " / " indicates that the previous and next associated objects are in a "division" relationship. "Including at least one of A, B and C" can mean: including A; including B; including C; including A and B; including A and C; including B and C; including A, B and C.

[0175] It is understood that the various numbers used in the embodiments of this application are merely for ease of description and are not intended to limit the scope of the embodiments of this application. The order of the sequence numbers of the above-mentioned processes does not necessarily imply a specific order of execution; the order of execution of the processes should be determined by their functions and inherent logic.

Claims

1. A neural network model training method, characterized in that: The method comprises: Get target feedback mode and target update mode; At a feedback time corresponding to the target feedback mode, receiving a feedback result for a forward reasoning result from a receiving device, wherein the forward reasoning result is obtained based on a first neural network model deployed in a sending device; Based on the target update mode and the feedback result, the model parameters of the first neural network model are updated.

2. The method according to claim 1, characterized in that: The acquisition target feedback mode and target update mode include: Acquire one or more of the first capability information of the transmitting end device, the second capability information of the receiving end device, or the first channel information, where the first channel information is information corresponding to the channel between the transmitting end device and the receiving end device; Determining a target feedback mode based on one or more of the first capability information, the second capability information, or the first channel information; A target update mode is determined based on one or more of the first capability information, the second capability information, or the first channel information.

3. The method according to claim 1 or 2, characterized in that: The target feedback mode is a first feedback mode, and the method further includes: Determine the feedback time according to the time interval and the sending time of the forward reasoning result; wherein the time interval is the interval between the sending time of the forward reasoning result and the feedback time of the feedback result, and the time interval is determined based on one or more of the first capability information, the second capability information or the first channel information.

4. The method according to claim 1 or 2, characterized in that: The target feedback mode is a second feedback mode, and the method further includes: Receive first indication information from the receiving end device, where the first indication information is used to indicate the feedback time.

5. The method according to claim 1 or 2, characterized in that: The target feedback mode is a second feedback mode, and the method further includes: receiving a request message from the receiving end device, wherein the request message is used to request sending a feedback result; A confirmation message is sent to the receiving end device, where the confirmation message includes first indication information, and the first indication information is used to indicate the feedback time.

6. The method according to claim 4 or 5, characterized in that: The number of the forward reasoning results is multiple, the number of the feedback results is multiple, the forward reasoning results and the feedback results correspond to each other one by one, and the method further includes: Receive second indication information from the receiving end device, where the second indication information is used to indicate a corresponding relationship between each feedback result and each forward reasoning result.

7. The method according to any one of claims 1 to 6, characterized in that: The target update mode is a first update mode or a second update mode; wherein the first update mode indicates updating after receiving N feedback results, and N is an integer greater than or equal to 1; the second update mode indicates updating after receiving third indication information, and the third indication information is used to indicate updating of model parameters of the first neural network model.

8. A neural network model training method, characterized in that: The method comprises: Get target feedback mode; At a feedback time corresponding to the target feedback mode, a feedback result for a forward reasoning result is sent to a sending end device; the forward reasoning result is obtained based on a first neural network model deployed in the sending end device.

9. The method according to claim 8, characterized in that: The obtaining target feedback mode includes: Acquire one or more of first capability information of the transmitting end device, second capability information of the receiving end device, or first channel information, where the first channel information is information corresponding to a channel between the transmitting end device and the receiving end device; The target feedback mode is determined according to one or more of the first capability information, the second capability information, or the first channel information.

10. The method according to claim 8 or 9, characterized in that: The target feedback mode is a first feedback mode, and the method further includes: The feedback time is determined according to the time interval and the sending time of the forward reasoning result; wherein the time interval is the forward The interval between the sending time of the inference result and the feedback time of the feedback result, wherein the time interval is determined based on one or more of the first capability information, the second capability information, or the first channel information.

11. The method according to claim 8 or 9, characterized in that: The target feedback mode is a second feedback mode, and the method further includes: Sending first indication information to the transmitting end device, where the first indication information is used to indicate the feedback time.

12. The method according to claim 8 or 9, characterized in that: The target feedback mode is a second feedback mode, and the method further includes: Sending a request message to the sending end device, wherein the request message is used to request sending a feedback result; A confirmation message is received from the sending end device, where the confirmation message includes first indication information, and the first indication information is used to indicate the feedback time.

13. The method according to claim 11 or 12, characterized in that: The number of the forward reasoning results is multiple, the number of the feedback results is multiple, the forward reasoning results and the feedback results correspond to each other one by one, and the method further includes: Sending second indication information to the sending end device, where the second indication information is used to indicate a corresponding relationship between each feedback result and each forward reasoning result.

14. The method according to any one of claims 8 to 13, characterized in that: The method further comprises: Obtain a target update mode, which is a first update mode or a second update mode; wherein the first update mode indicates updating after receiving N feedback results, and N is an integer greater than or equal to 1; the second update mode indicates updating after receiving third indication information, and the third indication information is used to indicate updating of model parameters of the first neural network model.

15. The method according to claim 14, characterized in that: The target update mode is the second update mode, and the method further includes: Send the third indication information to the sending end device.

16. The method according to claims 10-15, characterized in that: The receiving end device is deployed with a second neural network model, and the method further includes: According to the forward reasoning result, the model parameters of the second neural network model are updated.

17. A communication device, characterized in that: The method comprises a module for executing the method according to any one of claims 1 to 7, or comprises a module for executing the method according to any one of claims 8 to 16.

18. A communication device, characterized in that: It includes a processor and an interface circuit, wherein the interface circuit is used to receive signals from other communication devices outside the communication device and transmit them to the processor or send signals from the processor to other communication devices outside the communication device, and the processor is used to implement the method as described in any one of claims 1 to 7 through a logic circuit or by executing code instructions, or the processor is used to implement the method as described in any one of claims 8 to 16 through a logic circuit or by executing code instructions.

19. The method according to claim 18, characterized in that: The communication device is a chip or a chip system.

20. A computer-readable storage medium, characterized in that: The storage medium stores a computer program or an instruction. When the computer program or the instruction is executed by the communication device, the method according to any one of claims 1 to 7 is implemented, or the method according to any one of claims 8 to 16 is implemented.

21. A computer program product, characterized in that The computer program product includes a computer program or instructions. When the computer program or instructions are executed by a communication device, the communication device implements the method according to any one of claims 1 to 7, or implements the method according to any one of claims 8 to 16.

Images