Model training method and communication apparatus

Abstract

Provided in the present application are a model training method and a communication apparatus. The method comprises: a first device receiving from a second device data feature information of a second local data set, the second local data set being a data set of the second device; on the basis of the data feature information of the second local data set and a first local data set, determining a first training data set, the first local data set being a data set of the first device, and data feature information of the first training data set comprising the data feature information of the second local data set and data feature information of the first local data set; and, on the basis of the first training data set, obtaining a first model, the first model being a model deployed in the first device. By means of the method, the first device can perform model training with reference to the local data set of the other device, aiding in subsequent model alignment with the other device and helping to shorten the time for the model alignment.

Description

A model training method and communication device

[0001] This application claims priority to the Chinese patent application with application number 202311737963.5 filed with the State Intellectual Property Office of China on December 15, 2023, and priority to the Chinese patent application with the invention name “A model training method and communication device”, all contents of which are incorporated by reference into this application. Technical Field

[0002] The present application relates to the field of communication technology, and in particular to a model training method and a communication device. Background Art

[0003] 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.

[0004] However, the transmitting and receiving devices during communication may belong to different manufacturers. This means that the model deployed on the transmitting device (referred to as the transmitting model) and the model deployed on the receiving device (referred to as the receiving model) correspond to different training processes, which may cause the transmitting and receiving devices to be unable to communicate. To avoid this, the transmitting and receiving models need to be aligned (also called adapted or jointly trained) before the transmitting and receiving devices communicate.

[0005] Typically, the alignment process of the sending and receiving models can be understood as training them using data from a single training dataset. However, since the training of the sending and receiving models is independent, the alignment training process is time-consuming and expensive. Summary of the Invention

[0006] The present application provides a model training method and a communication device, which can refer to the local data sets of other devices for model training, which is conducive to subsequent model alignment with other devices, thereby shortening the time of model alignment.

[0007] In a first aspect, the present application provides a model training method, which is applied to a first device, or to a module in the first device (such as a chip or a chip system, etc.). Taking the application to the first device as an example, the method includes: the first device receives data feature information of a second local data set from a second device, and the second local data set is a data set of the second device; further, based on the data feature information of the second local data set and the first local data set, a first training data set is determined, and the first local data set is a data set of the first device, and the data feature information of the first training data set includes the data feature information of the second local data set and the data feature information of the first local data set; then, the first device obtains a first model based on the first training data set, and the first model is a model deployed on the first device. Optionally, the data feature information of the second local data set is indicated by first indication information, that is, the first device receives first indication information from the second device, and the first indication information is used to indicate the data feature information of the second local data set.

[0008] In one possible implementation method, a first device receives a second local data set from a second device, where the second local data set is the data set of the second device; further, the first device determines a first training data set based on the second local data set and the first local data set, where the first local data set is the data set of the first device; and further, the first device obtains a first model based on the first training data set, where the first model is the model deployed on the first device.

[0009] Based on the method described in the first aspect, before obtaining the first model deployed on the first device, the first device obtains a first training data set with reference to the data feature information of the second local data set of the second device, and further obtains the first model based on the first training data set. Through such a method, in terms of feature representation, the first model and the model obtained based on the second local data set (e.g., the second model) can be similar, which is conducive to establishing a connection between the model of the first device (i.e., the first model) and the model of the second device (i.e., the second model), thereby reducing the complexity of subsequent model alignment between the first model and the second model, and shortening the time of model alignment.

[0010] In one possible embodiment, the data feature information includes one or more of the following information: data identification information, data distribution information, or data classification information; wherein the data identification information is used to indicate the data included in the data set, the data distribution information is used to indicate the distribution of the data in the data set, and the data classification information is used to indicate the classification of the data in the data set.

[0011] In one possible embodiment, the data distribution information includes one or more of cluster distribution information, probability distribution information, or model parameter information of a data generation model; the data classification information includes one or more of geographic area information, signal-related information, device configuration information, quality-related information, or classification center information.

[0012] In one possible implementation, the first device determines data feature information of a second data set difference set from the data feature information of the second local data set based on the data feature information of the first local data set, where the data feature information of the second data set difference set is data feature information not included in the first local data set. Furthermore, the first device sends the data feature information of the second data set difference set to the second device and receives the second data set difference set from the second device. The first training data set is the union of the second data set difference set and the first local data set. Optionally, the data feature information of the second data set difference set is indicated by second indication information, that is, the first device sends second indication information to the second device, where the second indication information is used to indicate the data feature information of the second data set difference set.

[0013] In one possible implementation, the first device aligns the first model with the second model, where the second model is a model deployed on the second device, and the second model is obtained based on a second training data set, where the data feature information of the first training data set includes the data feature information of the second training data set. By implementing this possible implementation, the data feature information of the first training data set includes the data feature information of the second training data set, and the first model and the second model obtained therefrom will have certain similarities in feature representation. Furthermore, in the process of model alignment, compared to aligning models that do not have similarity, the present application helps to improve the speed of model alignment by aligning models that have certain similarities.

[0014] Optionally, the second training dataset includes data feature information of a second local dataset, or the second training dataset includes data feature information of the second local dataset and data feature information of the first local dataset. If the second training dataset includes data feature information of the second local dataset and data feature information of the first local dataset, the data feature information of the second training dataset is the same as the data feature information of the first training dataset.

[0015] In one possible implementation, the first device determines data feature information of a first data set difference set from the first local data set based on data feature information of the second local data set, where the data feature information of the first data set difference set is data feature information not included in the second local data set. Furthermore, the first device sends the first data set difference set to the second device. Optionally, the first data set difference set is indicated by third indication information, that is, the first device sends third indication information to the second device, where the third indication information is used to indicate the first data set difference set.

[0016] In one possible implementation, the first device determines a reference model based on the size of the first training data set and the size of the second training data set, where the reference model is the first model or the second model; further, the first device aligns the first model and the second model based on the reference model.

[0017] In a possible implementation, the data feature information included in the first training data set further includes data feature information of a third training data set. The third training data set is used to train a third model, and the third model is a model deployed on a third device.

[0018] In a second aspect, the present application provides a data transmission method, which is applied to a second device, or to a module (e.g., a chip or chip system) in the second device. Taking application to the second device as an example, the method includes: the second device sending data characteristic information of a second local data set to a first device, where the second local data set is a data set of the second device. Optionally, the data characteristic information of the second local data set is indicated by first indication information, that is, the second device sends first indication information to the first device, where the first indication information is used to indicate the data characteristic information of the second local data set.

[0019] In one possible implementation method, the second device sends a second local data set to the first device.

[0020] Based on the method described in the second aspect, the second device can provide its own local data set (i.e., the second local data set) to other devices (e.g., the first device) for reference, which is conducive to establishing a connection between the model of the first device (i.e., the first model) and the model of the second device (i.e., the second model), thereby helping to reduce the complexity of subsequent model alignment between the first model and the second model, and helping to shorten the time for model alignment. Based on the beneficial effects of other implementations described in the second aspect, reference can be made to the beneficial effects of the implementation described in the first aspect, and no further details will be given later.

[0021] In one possible embodiment, the data feature information includes one or more of the following information: data identification information, data distribution information, or data classification information; wherein the data identification information is used to indicate the data included in the data set, the data distribution information is used to indicate the distribution of the data in the data set, and the data classification information is used to indicate the classification of the data in the data set.

[0022] In one possible embodiment, the data distribution information includes one or more of cluster distribution information, probability distribution information, or model parameter information of a data generation model; the data classification information includes one or more of geographic area information, signal-related information, device configuration information, quality-related information, or classification center information.

[0023] In one possible implementation, the second device receives data feature information of a second data set difference set from the first device, where the data feature information of the second data set difference set is data feature information included in the second local data set and not included in the first local data set, where the first local data set is the data set of the first device. Furthermore, based on the data feature information of the second data set difference set, the second device sends the second data set difference set to the first device. Optionally, the data feature information of the second data set difference set is indicated by second indication information, that is, the second device receives second indication information from the first device, where the second indication information is used to indicate the data feature information of the second data set difference set.

[0024] In one possible implementation, the second device aligns the first model with the second model, where the second model is a model deployed on the second device. The second model is obtained based on a second training data set, and the data feature information of the first training data set includes the data feature information of the second training data set.

[0025] Optionally, the second training dataset includes data feature information of a second local dataset, or the second training dataset includes data feature information of the second local dataset and data feature information of the first local dataset. If the second training dataset includes data feature information of the second local dataset and data feature information of the first local dataset, the data feature information of the second training dataset is the same as the data feature information of the first training dataset.

[0026] In one possible implementation, the second device receives a first dataset difference set from the first device, where the data feature information of the first dataset difference set is data feature information included in the first local dataset and not included in the second local dataset; and the second training dataset is the union of the first dataset difference set and the second local dataset. Optionally, the first dataset difference set is indicated by third indication information, that is, the second device receives third indication information from the first device, where the third indication information is used to indicate the first dataset difference set.

[0027] In one possible implementation, the second device determines a reference model based on the size of the first training data set and the size of the second training data set, where the reference model is the first model or the second model; further, the second device aligns the first model and the second model based on the reference model.

[0028] In a possible implementation, the data feature information included in the first training data set further includes data feature information of a third training data set. The third training data set is used to train a third model, and the third model is a model deployed on a third device.

[0029] In a third aspect, the present application provides a communication device, which may be a first device, a device in the first device, or a device that can be used in conjunction with the first 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. The hardware or software includes one or more units or modules corresponding to the above functions. The unit or module 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.

[0030] In a fourth aspect, the present application provides a communication device, which may be a second device, a device in the second device, or a device that can be used in conjunction with the second 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.

[0031] 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.

[0032] In a sixth aspect, the present application provides a communication device, comprising a processor connected to a memory, configured to call a program stored in the memory to execute the method described in the first or second aspect above. The memory may be located within or outside the first or second device. The processor may include one or more processors.

[0033] In the seventh aspect, the present application provides a computer-readable storage medium, which stores a computer program or instruction. When the computer program or instruction is executed by a communication device, it implements the method described in the first aspect, or implements the method described in the second aspect.

[0034] In an eighth 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.

[0035] In a ninth 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

[0036] FIG1a is a schematic diagram of the architecture of a communication system provided in an embodiment of the present application;

[0037] FIG1b is another schematic diagram of a wireless communication system applicable to an embodiment of the present application;

[0038] 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;

[0039] 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;

[0040] FIG4 is a flow chart of a model training method provided in an embodiment of the present application;

[0041] FIG5 is a schematic diagram of classification of a second local data set provided in an embodiment of the present application;

[0042] FIG6 is a schematic diagram of a relationship between a first local data set and a second local data set provided in an embodiment of the present application;

[0043] FIG7 is a schematic diagram of a flow chart of a model alignment method provided in an embodiment of the present application;

[0044] FIG8 is a schematic diagram of a model alignment provided in an embodiment of the present application;

[0045] FIG9 is a schematic diagram of another model alignment provided in an embodiment of the present application;

[0046] FIG10 is a schematic diagram of another model alignment provided in an embodiment of the present application;

[0047] FIG11 is a schematic structural diagram of a communication device provided in an embodiment of the present application;

[0048] FIG12 is a schematic structural diagram of another communication device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0049] 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.

[0050] 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.

[0051] 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).

[0052] 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.

[0053] 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.

[0054] 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.

[0055] 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., resource element (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.

[0056] 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.

[0057] 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.

[0058] 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.

[0059] 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.

[0060] 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.

[0061] 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.

[0062] 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.

[0063] Please refer to FIG. 1 b , which is another schematic diagram of a wireless communication system applicable to an embodiment of the present application.

[0064] As shown in Figure 1b, the wireless communication system includes a RAN intelligent controller (RIC). As an example, the RIC can be used to implement functions related to artificial intelligence (AI). 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. The 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.

[0065] 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.

[0066] 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.

[0067] The near-real-time RIC and non-real-time RIC may also be separately configured as network elements. 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 may be configured in a RAN node (e.g., a CU or DU), while the non-real-time RIC may be configured in an operation, administration, and maintenance (OAM) system, a cloud server, a core network device, or other network devices.

[0068] 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.

[0069] 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.

[0070] 1. Neural Networks

[0071] 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).

[0072] Where, s = 1, 2, ... n, n is a natural number greater than 1, W s is x sThe 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.

[0073] 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.

[0074] 2. Intelligent air interface technology

[0075] 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.

[0076] 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.

[0077] In 3a of Figure 3, after the transmitted bits are encoded by the forward error correction (FEC) of the transmitting device 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 through 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 of the receiving device, transformed into frequency domain signals through fast Fourier transform (FFT), and then demodulated by the demodulation neural network model (i.e., the NN-DeMod module in 3a of Figure 3) and sent to FEC decoding.

[0078] In 3b of Figure 3, the bits to be transmitted are processed by the coding modulation neural network model of the transmitting device (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.

[0079] In 3c of Figure 3, the transmitting device 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 device 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.

[0080] In the application process of intelligent air interface technology, there are situations where the model deployed on the transmitting device (referred to as the transmitting model for short) and the model deployed on the receiving device (referred to as the receiving model for short) correspond to different training processes. For example, the transmitting device and the receiving device belong to different manufacturers. In this case, if the transmitting device and the receiving device do not align the transmitting model and the receiving model before communication, it may happen that the receiving model cannot parse the data processed by the transmitting model, resulting in the inability to communicate between the transmitting device and the receiving device. Generally, the model alignment process of the transmitting model and the receiving model is a process of aligning the transmitting model and the receiving model with the data in a training data set (or understood as joint training, adaptation training, etc.). The model alignment process takes a long time and has a high overhead.

[0081] This application provides a model training method that can improve the correlation between the training data set corresponding to the sending model and the training data set corresponding to the receiving model, thereby facilitating the correlation between the sending model and the receiving model and facilitating subsequent model alignment. The model training method and communication device provided by this application are further described below with reference to the accompanying drawings:

[0082] Please refer to Figure 4, which is a flow chart of a model training method provided by an embodiment of the present application. As shown in Figure 4, the model training method includes the following S401 to S403. The execution subject of the method shown in Figure 4 can be the first device and the second device, or the execution subject of the method shown in Figure 4 can be a module in the first device and a module in the second device, or the execution subject of the method shown in Figure 4 can be a chip of the first device and a chip of the second device. Figure 4 is illustrated by taking the first device and the second device as the execution subject of the method as an example. It should be noted that the first device (or second 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 limitations. It should also be noted that the first device in this application can be a sending device or a receiving device in the communication process, and the second device in this application can also be a sending device or a receiving device; when the first device is a sending device, the second device is a receiving device; when the first device is a receiving device, the second device is a sending device. Wherein:

[0083] S401: The second device sends data feature information of a second local data set to the first device. The second local data set is a data set of the second device.

[0084] It can be understood that in this application, one or more of the data set collected by the first device, the data set generated by the data generation model, or the stored data set is recorded as the first local data set, and the first local data set is the data set of the first device; the data set collected by the second device, the data set generated by the data generation model, or the stored data set is recorded as the second local data set, and the second local data set is the data set of the second device. After the second device obtains the data feature information of the second local data set, the second device indicates the data feature information of the second book data set to the first device. In the following text, this application is explained by taking the example of the second device indicating the data feature information of the second local data set to the first device through the first indication information.

[0085] The data characteristic information is used to indicate the characteristics of the data included in the second local data set. This application does not specifically limit the specific form of the data characteristic information. For ease of understanding, this application provides several types of data characteristic information for exemplary illustration, which should not be regarded as a specific limitation of this application. In one possible implementation method, the data characteristic information includes one or more of data identification information, data distribution information, or data classification information.

[0086] ①. Data identification information is used to indicate the data included in the data set. The data identification information may be an identity document (ID) of the data in the data set. In this case, the first indication information includes the IDs of all data in the second local data set.

[0087] ② Data distribution information (or data statistical information) is used to indicate the distribution of data in the dataset. The data distribution information can be cluster distribution information (e.g., the number of clusters corresponding to the second local dataset, the cluster centers of each cluster, the cluster sizes of each cluster, or the cluster densities of each cluster, etc.), probability distribution information (used to indicate the parameters of the probability distribution satisfied by the data in the dataset, e.g., if the data in the second local dataset satisfies a Gaussian distribution, the probability distribution information is the mean and variance of the Gaussian distribution), or model parameter information of a data generation model (used to indicate the model parameters that generated the data if the data in the dataset was generated by a data generation model).

[0088] ③. Data classification information is used to indicate the classification of data in a data set. Among them, the data classification information includes information about the categories included in the data set, the size of the subsets corresponding to each category (that is, the number of data included in each subset), or one or more of the data included in the subsets corresponding to each category. It should be noted that this application does not specifically limit the category information included in the data set. The category information can be one or more of geographic area information (such as cell ID, site ID, longitude and latitude information of the data acquisition device, etc.), signal-related information (such as Doppler spread information, delay spread information or antenna correlation coefficient corresponding to the signal when collecting data), device configuration information (such as the moving speed or antenna configuration information of the device collecting data), quality-related information (such as data processing algorithm-related information when obtaining data, channel estimation accuracy information when collecting data, etc.), and classification center information (such as cluster distribution information corresponding to the data set).

[0089] For example, based on Doppler spread information and delay spread information, the multiple signals included in the second local data set are classified to obtain four subsets (i.e., subsets 1 to 4) as shown in Figure 5. Subset 1 corresponds to Doppler spread range 1 and delay spread range 1, subset 2 corresponds to Doppler spread range 1 and delay spread range 2, subset 3 corresponds to Doppler spread range 2 and delay spread range 1, and subset 4 corresponds to Doppler spread range 2 and delay spread range 2. Furthermore, the second device sends first indication information to the first device, where the first indication information indicates the Doppler spread range and delay spread range corresponding to each of subsets 1 to 4 included in the second local data set, or the first indication information indicates the quantity or proportion of data included in each subset of the second local data set.

[0090] It should be noted that this application does not specifically limit the manner in which a device obtains data feature information of its local dataset. Specifically, this application does not limit the manner in which a second device obtains data feature information of a second local dataset, nor does it limit the manner in which a first device obtains data feature information of a first local dataset. For example, the second device may obtain the data feature information of the second local dataset by analyzing the second local dataset using a data feature information analysis model (e.g., a clustering algorithm). Alternatively, the second device may obtain the data feature information of the second local dataset from the attribute information of the second local dataset (which may be understood as including information describing the local dataset) without analyzing the second local dataset using a data feature information analysis model.

[0091] It should also be noted that the devices in this application have a consensus on the data feature information (i.e., the understanding is consistent). For example, when the first indication information indicates that the data feature information of the second local data set is data identification information, the data feature information of the first local data set of the first device is also data identification information; when the first indication information indicates that the data feature information of the second local data set is data distribution information, and the data distribution information is cluster distribution information, the data feature information of the first local data set of the first device is also cluster distribution information; when the first indication information indicates that the data feature information of the second local data set is data classification information, and the second local data set is classified according to Doppler spread range 1 and delay spread range 1, the data feature information of the first local data set of the first device will also be classified according to Doppler spread range 1, other Doppler spread ranges (ranges other than Doppler spread range 1), delay spread range 1, and other delay spread ranges (i.e., ranges other than delay spread range 1).

[0092] In a possible implementation, before S401 is executed, the first device and the second device may perform a data set alignment negotiation (or understand as starting a data set alignment process, ie, triggering the execution of S401 to S402).

[0093] For example, in one possible implementation, the first device sends a dataset alignment request message to the second device; the second device, based on the dataset alignment request message, sends a dataset alignment confirmation message to the first device; further, the second device sends the first indication information to the first device. The dataset alignment confirmation message may be the same message as the message carrying the first indication information in S401, or may be a different message from the message carrying the first indication information in S401.

[0094] For another example, in another possible implementation, the second device sends a data set alignment request message to the first device; the first device sends a data set alignment confirmation message to the second device based on the data set alignment request message; further, the second device sends the first indication information to the first device.

[0095] S402. The first device determines a first training data set based on the data feature information of the second local data set and the first local data set. The first local data set is a data set of the first device, and the data feature information of the first training data set includes the data feature information of the second local data set and the data feature information of the first local data set.

[0096] That is, the first device combines the data feature information of the first local dataset and the data feature information of the second local dataset to obtain a first training dataset. The first training dataset may be a union of the first local dataset and the second local dataset, or the data feature information of the first training dataset includes the data feature information of the first local dataset and the data feature information of the second local dataset.

[0097] In one possible implementation of S402, the first device generates a data set (referred to as a first generated data set) having the same data feature information as the second local data set based on the data feature information of the second local data set. Furthermore, the first device determines a first training data set based on the first generated data set and the first local data set, where the first training data set is the union of the first generated data set and the first local data set.

[0098] For example, the data feature information of the second local dataset is model parameter information for data generation model 1. In this case, the first device constructs data generation model 1 based on the model parameter information and generates a first generated dataset using data generation model 1. Furthermore, the first device determines the union of the first generated dataset and the first local dataset as the first training dataset. In other words, the data feature information of the first training dataset can be considered to include the data feature information of the first local dataset and the data feature information of the second local dataset.

[0099] In another possible implementation of S402, the first device determines data feature information of a second data set difference set from the second local data set based on the data feature information of the first local data set, where the data feature information of the second data set difference set is data feature information not included in the first local data set. Furthermore, the first device sends the data feature information of the second data set difference set to the second device, receives the second data set difference set from the second device, and determines a first training data set based on the second data set difference set and the first local data set. The first training data set is the union of the first local data set and the second data set difference set. Hereinafter, this application will be described using the example of the first device indicating the data feature information of the second data set difference set to the second device via second indication information.

[0100] For ease of understanding, as shown in FIG6 , there is an intersection between the data feature information of the first local dataset and the data feature information of the second local dataset. This intersection can be an empty set or a non-empty set, and this application does not specifically limit this. In this application, the difference between the data feature information of the first local dataset and the intersection is referred to as the data feature information of the first dataset difference set; the difference between the data feature information of the second local dataset and the intersection is referred to as the data feature information of the second dataset difference set. After receiving the data feature information of the second local dataset, the first device compares the data feature information of the first local dataset with the data feature information of the second local dataset, and determines from the data feature information of the second local dataset that data feature information that is different from the data feature information of the first local dataset (i.e., the data feature information of the second dataset difference set). Furthermore, the first device sends second indication information to the second device, where the second indication information is used to indicate the data feature information of the second dataset difference set. Based on the data feature information indicated by the second indication information, the second device sends the second dataset difference set to the first device. Furthermore, the first device determines the union of the second dataset difference set and the first local dataset as the first training dataset.

[0101] S403: The first device obtains a first model based on the first training data set, where the first model is a model deployed on the first device.

[0102] That is, the first device trains the neural network model based on the first training data set to obtain a first model. The first model can be a model deployed in a transmitter of the first device or a model deployed in a receiver of the first device.

[0103] In summary, the first device obtains a first training data set with reference to the data feature information of the second local data set of the second device, and obtains a first model based on the first training data set. It is understandable that the first model and the model obtained based on the second local data set (e.g., the second model) have similarities in feature representation, which is conducive to establishing a connection between the model of the first device (i.e., the first model) and the model of the second device (i.e., the second model), thereby reducing the complexity of subsequent model alignment between the first model and the second model, and shortening the time for model alignment.

[0104] Optionally, the present application also provides a specific application scenario applicable to FIG4 , in which the first device may be an access network device, and the second device is a terminal device provided with services by the first device. Alternatively, the first device is a terminal device, and the second device is an access network device providing services to the first device.

[0105] In order to facilitate understanding of the model alignment process mentioned in this application, please refer to Figure 7, which is a flow chart of a model alignment method provided in an embodiment of the present application. As shown in Figure 7, the model alignment method includes the following S701 to S703. The execution subject of the method shown in Figure 7 can be the first device and the second device, or the execution subject of the method shown in Figure 7 can be a module in the first device and a module in the second device, or the execution subject of the method shown in Figure 7 can be a chip of the first device and a chip of the second device. Figure 7 is illustrated by taking the first device and the second device as the execution subject of the method as an example. It should be noted that the description of the first device and the second device can be referred to the description of the first device and the second device in the aforementioned Figure 4, which will not be repeated here. Among them:

[0106] S701: A first device obtains a first model based on a first training data set.

[0107] The description of S701 can be found in the description of the specific implementation of the aforementioned S401-S403, and will not be repeated here.

[0108] S702: The second device obtains a second model based on a second training data set.

[0109] That is, the second device trains the neural network model based on the second training data set to obtain a second model. The second model is a model deployed on the second device, for example, the second model is a model deployed in a transmitter of the second device, or a model deployed in a receiver of the second device.

[0110] The data feature information of the first training data set includes the data feature information of the second training data set. It should be noted that for the description of the data feature information and the first training data set, please refer to the relevant description of the data feature information and the first training data set in S401 to S403 above, which will not be repeated here.

[0111] For ease of understanding, the relationship between the first training data set and the second training data set is described below in two cases.

[0112] Case 1: The second training data set is the second local data set.

[0113] That is, the second training dataset includes the data feature information of the second local dataset. In this case, the data feature information of the first training dataset includes other data feature information in addition to the data feature information of the second training dataset. For example, as shown in FIG6 , the data feature information of the first training dataset also includes the data feature information of the first dataset difference set.

[0114] In a possible implementation, the data feature information of the first training data set includes not only the data feature information of the second training data set, but also the data feature information of a third training data set. The third training data set is used to train a third model, which is a model deployed on a third device.

[0115] Exemplarily, the central node maintains a first local dataset, distributed node 1 maintains a second local dataset, and distributed node 2 maintains a third local dataset. The central node sends data feature information of the first local dataset to distributed node 1 and distributed node 2. Based on the data feature information of the first local dataset and the data feature information of the second local dataset, distributed node 1 sends indication information 1 to the central node. Indication information 1 is used to indicate a second dataset difference set, which is composed of data corresponding to the data feature information included in the second local dataset but not included in the first local dataset. Based on the data feature information of the first local dataset and the data feature information of the third local dataset, distributed node 2 sends indication information 2 to the central node. Indication information 2 is used to indicate a third dataset difference set, which is composed of data corresponding to the data feature information included in the third local dataset but not included in the first local dataset. Further, the central node determines the union of the first local dataset, the second dataset difference set, and the third dataset difference set as a first training dataset for obtaining a first model deployed on the central node. Distributed node 1 determines the second local dataset as a second training dataset for obtaining a second model deployed on distributed node 1. Distributed node 2 determines the third local dataset as a third training dataset, used to obtain a third model deployed on distributed node 2. This means that the data feature information of the first training dataset includes the feature information of the second training dataset and the feature information of the third training dataset. Optionally, in one application scenario of this example, the central node may be an access network device or a core network device, and the multiple distributed nodes may be different terminal devices.

[0116] Case 2: The second training data set is the union of the second local data set and the difference set of the first data set.

[0117] In this case, the data feature information of the first training data set is the same as the data feature information of the second training data set. That is, the second training data set includes the data feature information of the second local data set and the data feature information of the first local data set. In one possible embodiment, after the first device obtains the data feature information of the second local data set, it determines the data feature information of the first data set difference from the first local data set based on the data feature information of the second local data set, where the data feature information of the first data set difference is data feature information included in the first local data set but not included in the second local data set; further, the first device sends the first data set difference to the second device; the second device determines the union of the first data set difference and the second local data set as the second training data set. Optionally, the first device indicates the first data set difference to the second device through third indication information.

[0118] For ease of understanding, as shown in FIG6 , the data feature information of the first training dataset is the union of the data feature information of the first local dataset and the data feature information of the difference set of the second dataset. The data feature information of the second training dataset is the union of the data feature information of the second local dataset and the data feature information of the difference set of the first dataset. In other words, the data feature information of the first training dataset and the data feature information of the second training dataset are the same.

[0119] S703: The first device and the second device align the first model and the second model.

[0120] That is, the first device and the second device align the first model and the second model (or understand it as alignment training, adaptation training, or joint training) based on the alignment data (which can be understood as training data for model alignment). It can be understood that the goal of aligning the first model and the second model is to enable the first device and the second device to communicate normally.

[0121] For example, in Example 1, the first model is the model of a transmitter deployed on a first device, and the second model is the model of a receiver deployed on a second device. After aligning the first and second models, the first model processes data #1 to obtain data #2; the first device sends data #2 to the second device; and the second model processes data #2 to obtain data #3. Data #3 is close to data #1, or the error between data #3 and data #1 is small (e.g., the error is less than a first threshold, which is not specified in this application).

[0122] For example, in Example 2, the first model is the model of the receiver deployed on the first device, and the second model is the model of the transmitter deployed on the second device. After aligning the first and second models, the second model processes data #1 to obtain data #2. The second device sends data #2 to the first device. The first model processes data #2 to obtain data #3. Data #3 is close to data #1, or the error between data #3 and data #1 is small.

[0123] In a possible implementation, aligning the first model and the second model includes but is not limited to the following three methods:

[0124] Method 1: Jointly train the first model and the second model.

[0125] The joint training process is shown in Figure 8. Where the first model is a model of a transmitter deployed on a first device (i.e., a transmitting device), and the second model is a model of a receiver deployed on a second device (i.e., a receiving device), the joint training process is shown in Figure 8a. Where the first model is a model of a receiver deployed on a first device (i.e., a receiving device), and the second model is a model of a transmitter deployed on a second device (i.e., a transmitting device), the joint training process is shown in Figure 8b.

[0126] The following takes 8a in Figure 8 as an example to illustrate the process of joint training. The first device performs forward reasoning on the aligned data through the first model to obtain a forward reasoning result, and sends the forward reasoning result to the second device; after the second device receives the forward reasoning result from the first device, it calculates the forward reasoning result through the second model (for example, including at least one of forward reasoning, loss function calculation or reverse gradient calculation), obtains a feedback result, and feeds back the feedback result to the first device. The feedback result can be the loss result of the second model reasoning calculation, or the gradient obtained by the second model performing reverse gradient calculation, and this application does not specifically limit this. The first device updates the model parameters of the first model based on the feedback result and the forward reasoning result. Optionally, the second device can also update the model parameters of the second model based on the forward reasoning result.

[0127] Method 2: Adaptive training is performed on the first model and the second model according to the third model in the adaptation node, where the adaptation node is the first device or the second device.

[0128] It should be understood that, when the adapter node is a first device, if the first model is the transmitter model in the first device and the second model is the receiver model in the second device, then the third model is the receiver model in the first device; if the first model is the receiver model in the first device and the second model is the transmitter model in the second device, then the third model is the transmitter model in the first device. When the adapter node is a second device, if the first model is the transmitter model in the first device and the second model is the receiver model in the second device, then the third model is the transmitter model in the second device; if the first model is the receiver model in the first device and the second model is the transmitter model in the second device, then the third model is the receiver model in the second device.

[0129] It should also be noted that this application does not specifically limit the method for determining the adapter node. For example, the first device and the second device determine the adapter node based on the size of the first training dataset and the size of the second training dataset. The adapter node can be a device with a larger training dataset or a device with a smaller training dataset, and this application does not specifically limit this. For ease of understanding, the following description uses the example of the adapter node being the device with the larger training dataset (i.e., the first device).

[0130] In one possible implementation of the second approach, the adapter node constructs a fourth training dataset based on the third model and sends the fourth training dataset to other devices (i.e., devices other than the adapter node in the first and second devices). This can be understood as the first and second devices aligning the first and second models based on the fourth training dataset.

[0131] For example, the first device is an adaptation node, and the first device trains the model of the transmitter and the model of the receiver in the first device based on the first training data set; the process of the adaptation training is shown in Figure 9. Wherein, the first model is the model of the transmitter deployed on the first device (i.e., the transmitting device), the third model is the model of the receiver deployed on the first device, and the second model is the model of the receiver deployed on the second device (i.e., the receiving device). The process of the adaptation training is shown in Figure 9a. Wherein, the first model is the model of the receiver deployed on the first device (i.e., the receiving device), the third model is the model of the transmitter deployed on the first device, and the second model is the model of the transmitter deployed on the second device (i.e., the transmitting device). The process of the adaptation training is shown in Figure 9b. Taking Figure 9a as an example, the process of the adaptation training is exemplified. The first device constructs a fourth training data set based on the input of the third model (i.e., the output of the first model) and the output of the third model, and sends the fourth training data set to the second device; the second device trains the second model based on the fourth training data set, so that when the input of the second model is the same as the input of the third model, the output of the second model is as close as possible to the output of the third model (or understood as the error is small).

[0132] In another possible implementation of the second approach, the adapter node constructs a training reference model based on the third model and transmits model parameters of the training reference model to other devices (i.e., devices other than the adapter node in the first and second devices). This can be understood as the first and second devices aligning the first and second models based on the training reference model.

[0133] For example, the first device is an adaptation node; the first device is a transmitting device, and the first device trains the transmitter model (i.e., the first model mentioned in this application) and the receiver model (i.e., the third model mentioned in this application) in the first device based on the first training data set; the second device is a receiving device, and the second device trains the receiver model (i.e., the second model mentioned in this application) in the second device based on the second training data set. In this case, the first device constructs a receiver training reference model based on the third model, and when the input of the receiver training reference model is the same as the input of the third model, the output of the receiver training reference model is the same or similar to the output of the third model. Furthermore, the first device sends the model parameters of the receiver training reference model to the second device; the second device constructs a fifth training data set based on the model parameters of the receiver training reference model, and trains the second model based on the fifth training data set, so that when the input of the second model is the same as the input of the receiver training reference model, the output of the second model is as close as possible to the output of the receiver training reference model.

[0134] Method 3: Aligning the first model and the second model based on a reference model, wherein the reference model is the first model or the second model, and the reference model is determined by the first device and the second device based on the size of the first training data set and the size of the second training data set.

[0135] It should be noted that the reference model can be a model obtained based on the larger training data set in the first model and the second model (i.e., the training data set corresponding to the larger numerical value between the first data amount and the second data amount); or, the reference model can also be a model obtained based on the smaller training data set in the first model and the second model (i.e., the training data set corresponding to the smaller numerical value between the first data amount and the second data amount); this application does not make specific restrictions on this.

[0136] That is, before aligning the first model and the second model, the size of the first training data set (recorded as the first data amount) and / or the size of the second training data set (recorded as the second data amount) are exchanged between the first device and the second device; then, the first device and / or the second device can determine the reference model based on the first data amount and the second data amount.

[0137] For example, in Example 3, the first device sends a first amount of data to the second device. After obtaining the first amount of data, the second device determines that the value of the first amount of data is greater than the value of the second amount of data. Furthermore, the second device determines the first model as the reference model. The second device sends fourth indication information to the first device, indicating that the reference model is the first model. Furthermore, the second device and the first device align the second model with the first model based on the first model. In Example 4, the second device sends a second amount of data to the first device. After obtaining the second amount of data, the first device determines that the value of the first amount of data is greater than the value of the second amount of data. Furthermore, the first device determines the first model as the reference model. The first device sends fifth indication information to the second device, indicating that the reference model is the first model. Furthermore, the second device and the first device align the second model with the first model based on the first model. In Example 5, the first device sends the first amount of data to the second device, and the second device sends the second amount of data to the first device. The first and second devices determine that the value of the first amount of data is greater than the value of the second amount of data. Furthermore, the first device and the second device determine the first model as a reference model. Furthermore, the second device and the first device align the second model with the first model based on the first model.

[0138] In a possible implementation of the third method, when the reference model is the first model, the second device trains the second model and the second adaptation layer based on the first model. For example, when the first device is a transmitting device and the second device is a receiving device, the first model is a model of the transmitter deployed on the first device, and the second model is a model of the receiver deployed on the second device, the process of the adaptation training is shown in 10a of Figure 10. The first device processes data #1 through the first model to obtain data #2; the first device sends data #2 to the second device; the second device processes data #2 through the second adaptation layer to obtain data #3, and processes data #3 through the second model to obtain data #4. The training goal of the adaptation training can be understood as making the error between data #4 and data #1 smaller. Or it can be understood as making the output of the joint processing of the second adaptation layer and the second model as close as possible to the output of the model of the receiver of the first device under the same input.

[0139] In another possible implementation of mode three, when the reference model is the second model, the first device trains the first model and the first adaptation layer based on the second model. For example, when the first device is a transmitting device and the second device is a receiving device, the first model is a model of the transmitter deployed on the first device, and the second model is a model of the receiver deployed on the second device, the process of the adaptation training is shown in 10b of Figure 10. The first device processes data #1 through the first model to obtain data #2, and processes data #2 through the first adaptation layer to obtain data #3; the first device sends data #3 to the second device; and the second device processes data #3 through the second model to obtain data #4. The training goal of the adaptation training can be understood as reducing the error between data #4 and data #1. Alternatively, it can be understood as making the output of the joint processing of the first model and the first adaptation layer as close as possible to the output of the transmitter model of the second device under the same input.

[0140] It can be seen that by implementing the model alignment method described in FIG. 7 of this application, the data feature information of the first training dataset includes the data feature information of the second training dataset, and the first model and the second model obtained based on this will have a certain similarity in feature representation. Furthermore, in the process of model alignment, compared to aligning models that do not have similarity, this application helps to improve the speed of model alignment by aligning models that have a certain degree of similarity.

[0141] It is understandable that, in order to realize the above functions, the above-mentioned devices include hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of each example described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0142] In the embodiments of the present application, the first device or the second device can be divided into functional modules according to the above method examples. For example, each functional module can be divided according to each function, or two or more functions can be integrated into one processing module. The above integrated modules can be implemented in the form of hardware or software functional modules. It should be noted that the division of modules in the embodiments of the present application is schematic and is only a logical functional division. In actual implementation, other division methods may be used.

[0143] Please refer to Figure 11, which shows a structural diagram of a communication device 1100 according to an embodiment of the present application. The communication device shown in Figure 11 can be a first device, or a device in the first device, or a device that can be used in combination with the first 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 second device, or a device in the second device, or a device that can be used in combination with the second 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, which may be data received by the communication unit 1101, and the processed data may also be sent by the communication unit 1101; the communication unit 1101 can be understood as a transceiver unit, including a receiving module and / or a sending module, the receiving module is used to perform the receiving action of the device (i.e., the first device or the second device) in any embodiment of Figure 4 or Figure 7, and the sending module is used to perform the sending action of the device (i.e., the first device or the second device) in any embodiment of Figure 4 or Figure 7.

[0144] In one embodiment, the communication device 1100 is a first device, a device in the first device (e.g., a chip or chip system in the first device), or a device that can be used in conjunction with the first device, wherein:

[0145] Communication unit 1101 is used to receive data feature information of a second local data set from a second device, where the second local data set is a data set of the second device; processing unit 1102 is used to determine a first training data set based on the data feature information of the second local data set and the first local data set, where the first local data set is a data set of the first device, and the data feature information of the first training data set includes the data feature information of the second local data set and the data feature information of the first local data set; processing unit 1102 is also used to obtain a first model based on the first training data set, where the first model is a model deployed on the first device.

[0146] In one possible implementation, the data feature information includes one or more of the following information: data identification information, data distribution information, or data classification information; wherein, the data identification information is used to indicate the data included in the data set, the data distribution information is used to indicate the distribution of the data in the data set, and the data classification information is used to indicate the classification of the data in the data set.

[0147] In one possible implementation, the data distribution information includes one or more of cluster distribution information, probability distribution information, or model parameter information of a data generation model; the data classification information includes one or more of geographic area information, signal-related information, device configuration information, quality-related information, or classification center information, Doppler information, or delay spread information.

[0148] In one possible implementation, the processing unit 1102 is further used to determine data feature information of a second data set difference set from the data feature information of the second local data set based on the data feature information of the first local data set, where the data feature information of the second data set difference set is data feature information not included in the first local data set; the communication unit 1101 is further used to send the data feature information of the second data set difference set to the second device; the communication unit 1101 is further used to receive the second data set difference set from the second device; the first training data set is the union of the second data set difference set and the first local data set.

[0149] In one possible implementation, the processing unit 1102 is further used to determine data feature information of a first data set difference set from the first local data set based on data feature information of the second local data set, where the data feature information of the first data set difference set is data feature information not included in the second local data set; and the communication unit 1101 is further used to send the first data set difference set to the second device.

[0150] In one possible implementation, the processing unit 1102 is further used to align the first model and the second model, where the second model is a model deployed on the second device, the second model is obtained based on the second training data set, and the data feature information of the first training data set includes the data feature information of the second training data set.

[0151] In a possible implementation, the data feature information of the second training data set is the same as the data feature information of the first training data set.

[0152] In one possible implementation, the processing unit 1102 is further used to determine data feature information of a first data set difference set from the first local data set based on data feature information of the second local data set, where the data feature information of the first data set difference set is data feature information not included in the second local data set; and the communication unit 1101 is further used to send the first data set difference set to the second device.

[0153] In one possible implementation, the processing unit 1102 is further used to determine a reference model based on the size of the first training data set and the size of the second training data set, where the reference model is the first model or the second model; the processing unit 1102 is further used to align the first model and the second model based on the reference model.

[0154] In a possible implementation, the data feature information included in the first training data set also includes data feature information of a third training data set. The third training data set is used to train a third model, and the third model is a model deployed on a third device.

[0155] For a more detailed description of the communication unit 1101 and the processing unit 1102 , reference may be made to the related description of the first device in the method embodiment shown in FIG. 4 or FIG. 7 .

[0156] In one embodiment, the communication device 1100 is a second device, a device in the second device, or a device that can be used in conjunction with the second device, wherein:

[0157] The communication unit 1101 is configured to send data feature information of a second local data set to the first device, where the second local data set is a data set of the second device.

[0158] In one possible implementation, the data feature information includes one or more of the following information: data identification information, data distribution information, or data classification information; wherein the data identification information is used to indicate the data included in the data set, the data distribution information is used to indicate the distribution of the data in the data set, and the data classification information is used to indicate the classification of the data in the data set.

[0159] In one possible implementation, the data distribution information includes one or more of cluster distribution information, probability distribution information, or model parameter information of a data generation model; the data classification information includes one or more of geographic area information, signal-related information, device configuration information, quality-related information, or classification center information.

[0160] In one possible implementation, the communication unit 1101 is further used to receive data feature information of a second data set difference set from the first device, where the data feature information of the second data set difference set is data feature information included in the second local data set and not included in the first local data set, and the first local data set is the data set of the first device; the communication unit 1101 is further used to send the second data set difference set to the first device based on the data feature information of the second data set difference set.

[0161] In one possible implementation, the processing unit 1102 is further used to align the first model and the second model, where the second model is a model deployed on the second device, and the second model is obtained based on a second training data set, and the data feature information of the first training data set includes the data feature information of the second training data set.

[0162] In a possible implementation, the data feature information of the second training data set is the same as the data feature information of the first training data set.

[0163] In one possible implementation, the communication unit 1101 is further used to receive a first data set difference set from the first device, where data feature information of the first data set difference set is included in the first local data set and does not include data feature information of the second local data set; the second training data set is the union of the first data set difference set and the second local data set.

[0164] In one possible implementation, the processing unit 1102 is further used to determine a reference model based on the size of the first training data set and the size of the second training data set, where the reference model is the first model or the second model; the processing unit 1102 is further used to align the first model and the second model based on the reference model.

[0165] In a possible implementation, the data feature information included in the first training data set also includes data feature information of a third training data set. The third training data set is used to train a third model, and the third model is a model deployed on a third device.

[0166] 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 second device in the method embodiment shown in FIG. 4 or FIG. 7 .

[0167] In a possible embodiment, when the communication device 1100 is a chip, the communication unit 1101 can be a communication interface, a pin or a circuit. The communication interface can be used to input data to be processed into the processor and can output the processing results of the processor. In a specific implementation, the communication interface can be a general purpose input and output (GPIO) interface, which can be connected to multiple peripheral devices (such as a display (LCD), a camera (camera), a radio frequency (RF) module, an antenna, etc.). The communication interface is connected to the processor via a bus.

[0168] The processing unit 1102 may be a processor that can execute computer-executable instructions stored in the storage module to cause the chip to perform the method described in any of the embodiments shown in FIG4 or FIG7. Furthermore, the processor may include a controller, an arithmetic unit, and registers. For example, the controller is primarily responsible for decoding instructions and issuing control signals for operations corresponding to the instructions. The arithmetic unit is primarily responsible for performing fixed-point or floating-point arithmetic operations, shift operations, and logical operations, and may also perform address operations and conversions. The registers are primarily responsible for storing register operands and intermediate operation results temporarily stored during instruction execution. In a specific implementation, the processor's hardware architecture may be an application-specific integrated circuit (ASIC) architecture, a microprocessor without interlocked piped stages architecture (MIPS) architecture, an advanced RISC machine (ARM) architecture, or a network processor (NP) architecture, among others. The processor may be single-core or multi-core. The storage module may be a memory module within the chip, such as a register or cache. The storage module may also be a storage module located outside the chip, such as a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM), etc.

[0169] It should be noted that the functions corresponding to the processor and the interface can be implemented through hardware design, software design, or a combination of hardware and software, and there is no limitation here.

[0170] Figure 12 is a schematic diagram of the structure of another communication device provided in an embodiment of the present application. It is understood that the communication device 1200 includes necessary means such as modules, units, elements, circuits, or interfaces, which are appropriately configured together to implement the present solution. The communication device 1200 can be the first device or the second device described above, or a component (such as a chip) in these devices, used to implement the method described in the above method embodiment.

[0171] In one possible design, as shown in FIG12 , the communication device 1200 includes a processor 1210 and an interface circuit 1220. The processor 1210 and the interface circuit 1220 are coupled to each other.

[0172] Optionally, the communication device 1200 may include one or more processors 1210. The processor 1210 may be a general-purpose processor or a dedicated processor. For example, it may be a baseband processor or a central processing unit. The baseband processor may be used to process communication protocols and communication data, and the central processing unit may be used to control the communication device (e.g., terminal equipment, network equipment, or chip), execute software programs, and process software program data.

[0173] It can be understood that the interface circuit 1220 can be a transceiver or an input / output interface. When the communication device 1200 is a first device or a second device, the interface circuit 1220 is a transceiver, including a transmitter and / or a receiver. Among them, the transmitter can be referred to as a transmitting unit, a transmitter or a transmitting circuit, etc., for implementing the transmitting function, and the receiver can be referred to as a receiving unit, a receiver or a receiving circuit, etc., for implementing the receiving function. When the communication device 1200 is a chip in the first device or the second device, the interface circuit 1220 is the input / output interface of the chip. Optionally, the communication device 1200 may also include an antenna (not shown in the figure), and the interface circuit 1220 may sometimes also be referred to as a transceiver unit, a transceiver, a transceiver circuit, or a transceiver, etc., for implementing the transceiver function of the communication device through the antenna.

[0174] Optionally, the communication device 1200 may further include a memory 1230 for storing instructions executed by the processor 1210, or storing input data required by the processor 1210 to execute instructions, or storing data generated after the processor 1210 executes instructions. Optionally, the processor 1210 and the memory 1230 may be provided separately or integrated together.

[0175] When the communication device 1200 is used to implement the method shown in FIG. 4 or FIG. 7 , 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 .

[0176] When the communication device is a chip implemented in a first device, the terminal chip implements the functions of the first device in the above-mentioned method embodiment. When the first device chip receives information from the second device, it can be understood that the information is first received by other modules in the first device (such as a radio frequency module or antenna) and then sent to the first device chip by these modules. When the first device chip sends information to the second device, it can be understood that the information is first sent to other modules in the first device (such as a radio frequency module or antenna) and then sent to the second device by these modules.

[0177] When the aforementioned communication device is a chip implemented in a second device, the second device chip implements the functions of the second device in the aforementioned method embodiment. When the second device chip receives information from the first device, it can be understood that the information is first received by other modules in the second device (such as a radio frequency module or antenna) and then transmitted to the second device chip by these modules. When the second device chip sends information to the first device, it can be understood that the information is transmitted to other modules in the second device (such as a radio frequency module or antenna) and then transmitted to the first device by these modules.

[0178] An embodiment of the present application further provides a computer-readable storage medium, which stores computer instructions. When the computer instructions are executed, the computer executes any of the methods described in any of the embodiments in FIG. 4 or FIG. 7 .

[0179] An embodiment of the present application further provides a computer program product, which includes: computer program code, and when the computer program code is executed by a computer, causes the computer to execute any of the methods described in any of the embodiments shown in FIG. 4 or FIG. 7 .

[0180] 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.

[0181] 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.

[0182] 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.

[0183] 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.

[0184] 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.

[0185] 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.

[0186] 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.

[0187] 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.

[0188] In this application, "sending" and "receiving" 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 transmission through the air interface, as well as indirect transmission through the air interface from other units or modules. "Receiving information from YY" can be understood as the source of the information being YY, which can include direct reception from YY through the air interface, as well as indirect reception 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. In other words, sending and receiving can be carried out between devices, for example, between a network device and a terminal device, or within a device, for example, between components, modules, chips, software modules, or hardware modules within the device through a bus, trace, or interface. It is understandable that information may undergo necessary processing, such as encoding, modulation, etc., between the source and destination of the information, 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.

[0189] The "indication" in this application 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, the information to be indicated can be directly indicated, such as the information to be indicated itself or the index of the information to be indicated. The information to be indicated can also be indirectly indicated by indicating other information, where there is an association between the other information and the information to be indicated; it is also possible to indicate only a part of the information to be indicated, while the other parts of the information to be indicated are known or agreed in advance. For example, the indication of specific information can be achieved with the help 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. This application does not limit the specific method of indication. It can be understood 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.

[0190] 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 model training method, characterized in that: The method comprises: Receiving data characteristic information of a second local data set from a second device, where the second local data set is a data set of the second device; Determine a first training data set based on the data feature information of the second local data set and the first local data set, where the first local data set is a data set of the first device, and the data feature information of the first training data set includes the data feature information of the second local data set and the data feature information of the first local data set; A first model is obtained based on the first training data set, where the first model is a model deployed on the first device.

2. The method according to claim 1, characterized in that: The determining the first training data set based on the data feature information of the second local data set and the first local data set includes: Based on the data feature information of the first local data set, determine the data feature information of a second data set difference set from the data feature information of the second local data set, where the data feature information of the second data set difference set is the data feature information not included in the first local data set; Sending data feature information of the second data set difference set to the second device; The second data set difference set is received from the second device; the first training data set is a union of the second data set difference set and the first local data set.

3. The method according to claim 1 or 2, characterized in that: The method further comprises: Aligning the first model with a second model, where the second model is a model deployed on the second device, the second model is obtained based on a second training data set, and data feature information of the first training data set includes data feature information of the second training data set; The second training data set includes data feature information of the second local data set, or the second training data set includes data feature information of the second local data set and data feature information of the first local data set.

4. The method according to claim 3, characterized in that: Before aligning the first model with the second model, the method further includes: determining data feature information of a first data set difference set from the first local data set according to the data feature information of the second local data set, where the data feature information of the first data set difference set is data feature information not included in the second local data set; The first data set difference set is sent to the second device, where the second training data set is a union of the first data set difference set and the second local data set.

5. A data transmission method, characterized in that: The method comprises: Data feature information of a second local data set is sent to the first device, where the second local data set is a data set of the second device.

6. The method according to claim 5, characterized in that: The method further comprises: receiving data feature information of a second data set difference set from the first device, where the data feature information of the second data set difference set is data feature information included in the second local data set and not included in the first local data set, where the first local data set is a data set of the first device; Based on the data feature information of the second data set difference set, the second data set difference set is sent to the first device.

7. The method according to claim 5 or 6, characterized in that: The method further comprises: Aligning the first model with a second model, where the second model is a model deployed on the second device, the second model is obtained based on a second training data set, and data feature information of the first training data set includes data feature information of the second training data set; The second training data set includes data feature information of the second local data set, or the second training data set includes data feature information of the second local data set and data feature information of the first local data set.

8. The method according to claim 7, characterized in that: Before aligning the first model with the second model, the method further includes: A first data set difference set is received from the first device, wherein data feature information of the first data set difference set is data feature information included in the first local data set and not included in the second local data set; and the second training data set is a union of the first data set difference set and the second local data set.

9. The method according to claim 3, 4, 7 or 8, characterized in that: The aligning the first model and the second model comprises: Determine a reference model based on the size of the first training data set and the size of the second training data set, where the reference model is the first model or the second model; The first model and the second model are aligned based on the reference model.

10. The method according to claim 3, 4, 7, 8 or 9, characterized in that: The data feature information included in the first training data set also includes data feature information of a third training data set, and the third training data set is used to train a third model, and the third model is a model deployed on a third device.

11. The method according to any one of claims 1 to 10, characterized in that: The data characteristic information includes one or more of the following information: data identification information, data distribution information or data classification information; The data identification information is used to indicate the data included in the data set, the data distribution information is used to indicate the distribution of the data in the data set, and the data classification information is used to indicate the classification of the data in the data set.

12. The method according to claim 11, characterized in that: The data distribution information includes one or more of cluster distribution information, probability distribution information, or model parameter information of a data generation model; The data classification information includes one or more of geographical area information, signal-related information, device configuration information, quality-related information or classification center information.

13. A communication device, characterized in that: Comprising modules for executing the method as claimed in any one of claims 1 to 12.

14. 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 12 through a logic circuit or executing code instructions.

15. The device according to claim 14, characterized in that: The communication device is a chip or a chip system.

16. 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 15 is implemented.

Images