Neural network training method and related device

Abstract

The invention discloses a neural network training method and a related device. The method comprises the steps that first equipment receives first channel sample information from second equipment; and the first device determines a first neural network, the first neural network is obtained by training according to the first channel sample information, and the first neural network is used for reasoning according to the first channel sample information to obtain second channel sample information. By means of the method, the air interface signaling overhead can be effectively reduced, meanwhile, the method can adapt to the located channel environment, and the communication performance is improved.

Description

[0001] This application is a divisional application. The application number of the original application is 202080107089.5, the original application date is December 31, 2020, and the entire content of the original application is incorporated herein by reference. Technical Field

[0002] This application relates to the field of communication technologies, and in particular, to a method for training a neural network and related devices. Background Art

[0003] A wireless communication system may include three parts: a transmitter, a channel, and a receiver. The channel is used to transmit signals exchanged between the transmitter and the receiver. Exemplarily, the transmitter may be an access network device, such as a base station (BS), and the receiver may be a terminal device. Again exemplarily, the transmitter may be a terminal device, and the receiver may be an access network device.

[0004] To optimize the performance of the communication system, the above-mentioned transmitter and receiver can be optimized. Both the transmitter and the receiver can have independent mathematical models. Therefore, usually, independent optimization is performed based on their respective mathematical models. For example, a mathematical channel model can be used to generate channel samples to optimize the transmitter and the receiver.

[0005] Since the mathematical channel model is non-ideal and non-linear, the channel samples generated by using the mathematical channel model defined by the protocol are difficult to reflect the real actual channel environment. When a large number of actual channel samples are transmitted between the transmitter and the receiver, it will occupy too much air interface resources and affect the data transmission efficiency. Summary of the Invention

[0006] A first aspect of an embodiment of this application provides a method for training a neural network, including:

[0007] A first device receives first channel sample information from a second device; the first device determines a first neural network, which is trained according to the first channel sample information and is used to infer second channel sample information.

[0008] Optionally, taking the first device as an access network device and the second device as a terminal device as an example for illustration. It can be understood that the first device may be an access network device, a chip used in the access network device, or a circuit used in the access network device, etc., and the second device may be a terminal device, a chip used in the terminal device, or a circuit used in the terminal device, etc.

[0009] In a possible design, the method includes: The first device trains a first neural network according to first channel sample information. The first neural network is used to generate new channel sample information, such as second channel sample information.

[0010] In a possible design, the method includes: The first device receives information of the first neural network from a third device and determines the first neural network according to the information of the first neural network. The first neural network is trained by the third device according to the first channel sample information.

[0011] In this method, the first neural network can be trained according to the first channel sample information. The first neural network is used to infer the second channel sample information. By this method, the air interface signaling overhead when transmitting channel sample information can be effectively reduced.

[0012] In a possible design, the second channel sample information is used to train a second neural network and / or a third neural network, and the second neural network and / or the third neural network are used for the first device to transmit target information to the second device.

[0013] In a possible design, the method includes: The first device trains the second neural network and / or the third neural network according to the second channel sample information, or according to the second channel sample information and the first channel sample information. The second neural network and / or the third neural network are used for the first device to transmit target information to the second device.

[0014] In a possible design, the method includes: The first device receives information of the second neural network and / or information of the third neural network from a third device. The second neural network and / or the third neural network are used for the first device to transmit target information to the second device. The second neural network and / or the third neural network are trained by the third device according to the second channel sample information, or according to the second channel sample information and the first channel sample information.

[0015] By this method, the air interface signaling overhead can be effectively reduced, and at the same time, it can adapt to the channel environment. The trained second neural network and third neural network are closer to the actual channel environment, improving the communication performance. The speed of training the second neural network and the third neural network is also greatly improved.

[0016] In a possible design, the method further includes: The first device sends a first reference signal to the second device. Optionally, the first reference signal includes a demodulation reference signal DMRS or a channel state information reference signal CSI-RS. Optionally, the sequence type of the first reference signal includes a ZC sequence or a Gold sequence.

[0017] In a possible design, the first channel sample information includes, but is not limited to, a second reference signal and / or channel state information (CSI). The second reference signal is the first reference signal that has propagated through the channel, or is described as the first reference signal received by the second device from the first device.

[0018] In a possible design, the information of the first neural network includes: the amount of change in the model of the first neural network relative to the reference neural network.

[0019] In a possible design, the information of the first neural network includes one or more of the following: the weights of the neural network, the activation function of the neurons, the number of neurons in each layer of the neural network, the inter-layer cascade relationship of the neural network, and the network type of each layer in the neural network.

[0020] In a possible design, the first neural network is a generative neural network. Optionally, the first neural network is a Generative Adversarial Network (GAN) or a Variational Autoencoder (VAE).

[0021] Through this method, second channel sample information that is identically distributed or has a similar distribution to the first channel sample information can be obtained, making the second channel sample information more closely approximate the actual channel environment.

[0022] In a possible design, the method further includes: the first device receiving, from the second device, the capability information of the second device. The capability information is used to indicate one or more of the following information:

[0023] 1) Whether the second device supports using a neural network to replace or implement the functions of the communication module;

[0024] The communication module includes, but is not limited to: an OFDM modulation module, an OFDM demodulation module, a constellation mapping module, a constellation demapping module, a channel coding module, a channel decoding module, a precoding module, an equalization module, an interleaving module, and / or a deinterleaving module.

[0025] 2) Whether the second device supports the network type of the third neural network;

[0026] 3) Whether the second device supports receiving the information of the third neural network through signaling;

[0027] 4) The reference neural network stored by the second device;

[0028] 5) The memory space available on the second device for storing the third neural network; and,

[0029] 6) Information on the computing power of the second device that can be used to run the neural network.

[0030] Through this method, the first device can receive the capability information sent by the second device. This capability information is used to notify the first device of relevant information about the second device, and the first device can perform operations on the third neural network based on this capability information to ensure that the second device can normally use this third neural network.

[0031] In a possible design, this method further includes: the first device sending information about the third neural network to the second device.

[0032] In this embodiment, after the first device finishes training the third neural network, the first device sends information about the third neural network to the second device. The information about the third neural network includes but is not limited to: the weights of the neural network, the activation function of the neurons, the number of neurons in each layer of the neural network, the inter-layer cascade relationship of the neural network, and / or the network type of each layer in the neural network. Exemplarily, different activation functions can be indicated for different neurons in the information about the third neural network.

[0033] In a possible design, when the third neural network is pre-configured (or pre-defined) in the second device, the information about the third neural network can also be the model change amount of the third neural network. The model change amount includes but is not limited to: the weights of the neural network that have changed, the activation function that has changed, the number of one or more layers of neurons in the neural network that have changed, the inter-layer cascade relationship of the neural network that has changed, and / or the network type of one or more layers in the neural network that have changed. Exemplarily, this pre-configuration can be pre-configured by the access network device through signaling, and this pre-definition can be pre-defined by the protocol. For example, the protocol pre-defines the third neural network in the terminal device as neural network A.

[0034] In the embodiments of this application, there can be multiple implementation schemes for the information about the third neural network, which improves the implementation flexibility of the scheme.

[0035] In a second aspect, the embodiments of this application provide a method for training a neural network, including: the second device performing channel estimation based on the first reference signal received from the first device to determine the first channel sample information; the second device sending the first channel sample information to the first device; the second device receiving information about the third neural network from the first device, where the third neural network is used for the second device and the first device to transmit target information.

[0036] In a possible design, this method further includes: the second device sending the capability information of the second device to the first device.

[0037] For the introduction of the first channel sample information, the information of the third neural network, and the capability information of the second device, etc., please refer to the first aspect and will not be elaborated here.

[0038] In a third aspect, an embodiment of the present application provides a method for training a neural network, including:

[0039] The first device sends a first reference signal to the second device; the first device receives the information of the first neural network from the second device; the first neural network is used to infer the second channel sample information.

[0040] In this method, the second device (such as a terminal device) trains the first neural network according to the first channel sample information. The first neural network is used to infer the second channel sample information. Through this method, the air interface signaling overhead when transmitting channel sample information can be effectively reduced.

[0041] In a possible design, the second channel sample information is used to train the second neural network and / or the third neural network, and the second neural network and / or the third neural network are used for the first device and the second device to transmit target information.

[0042] In a possible design, the method includes: the first device trains the second neural network and / or the third neural network according to the second channel sample information, or according to the second channel sample information and the first channel sample information, and the second neural network and / or the third neural network are used for the first device and the second device to transmit target information.

[0043] In a possible design, the method includes: the first device receives the information of the second neural network and / or the information of the third neural network from the third device, and the second neural network and / or the third neural network are used for the first device and the second device to transmit target information. The second neural network and / or the third neural network are trained by the third device according to the second channel sample information, or according to the second channel sample information and the first channel sample information.

[0044] Through this method, the air interface signaling overhead can be effectively reduced, and at the same time, it can adapt to the channel environment. The trained second neural network and third neural network are closer to the actual channel environment, improving the communication performance. The speed of training the second neural network and the third neural network is also greatly improved.

[0045] Specifically, for the introduction of the first reference signal, the first neural network, the information of the first neural network, the second neural network and / or the third neural network, etc., please refer to the first aspect and will not be elaborated.

[0046] In a possible design, the method further includes: the first device sends the information of the third neural network to the second device.

[0047] Specifically, for the introduction of the information of the third neural network, reference can be made to the first aspect, which will not be elaborated here.

[0048] In a possible design, the method further includes:

[0049] The first device receives capability information from the second device, where the capability information is used to indicate one or more of the following information of the second device:

[0050] 1) Whether it supports using a neural network to replace or implement the functions of the communication module;

[0051] 2) Whether it supports the network type of the first neural network;

[0052] 3) Whether it supports the network type of the third neural network;

[0053] 4) Whether it supports receiving the information of the reference neural network used to train the first neural network through signaling;

[0054] 5) Whether it supports receiving the information of the third neural network through signaling;

[0055] 6) The stored reference neural network;

[0056] 7) The memory space available for storing the first neural network and / or the third neural network;

[0057] 8) The computing power information available for running the neural network; and,

[0058] 9) The location information of the second device.

[0059] In a fourth aspect, an embodiment of the present application proposes a method for training a neural network, including:

[0060] The second device performs channel estimation based on the first reference signal received from the first device to determine first channel sample information; the second device determines a first neural network, where the first neural network is trained based on the first channel sample information; the second device sends the information of the first neural network to the first device.

[0061] Specifically, for the introduction of the first reference signal, the first neural network, the first channel sample information, and the information of the first neural network, reference can be made to the third aspect, which will not be elaborated here.

[0062] In a possible design, the method further includes:

[0063] The second device receives the information of the third neural network from the first device. For the information of the third neural network, reference can be made to the third aspect, which will not be elaborated here.

[0064] In a possible design, the method further includes: the second device sends capability information to the first device. For the capability information, please refer to the third aspect and details are not described herein.

[0065] In a fifth aspect, an embodiment of the present application provides a method for training a neural network, including: the third device receives first channel sample information from the first device; the third device trains a first neural network according to the first channel sample information, and the first neural network is used to infer second channel sample information.

[0066] In a possible design, the method further includes: training a second neural network and / or a third neural network according to the second channel sample information, and sending information of the second neural network and / or the third neural network to the first device, where the second neural network and / or the third neural network are used for the first device and the second device to transmit target information.

[0067] In a sixth aspect, a device is provided. The device may be an access network device, or a device in the access network device, or a device that can be used in combination with the access network device.

[0068] In a possible design, the device may include modules corresponding one by one to the methods / operations / steps / actions described in the first aspect. The module may be a hardware circuit, software, or a combination of hardware circuit and software. In one design, the device may include a processing module and a transceiver module. Exemplarily,

[0069] The transceiver module is used to receive first channel sample information from the second device;

[0070] The processing module is used to determine a first neural network, which is trained according to the first channel sample information and is used to infer second channel sample information.

[0071] For the introduction of the first neural network, the first channel sample information, the second channel sample information, and other operations, please refer to the first aspect and details are not described herein.

[0072] In a possible design, the device may include modules corresponding one by one to the methods / operations / steps / actions described in the third aspect. The module may be a hardware circuit, software, or a combination of hardware circuit and software. In one design, the device may include a processing module and a transceiver module. Exemplarily,

[0073] The transceiver module is used to send a first reference signal to the second device; and receive information of the first neural network from the second device; the first neural network is used to infer second channel sample information.

[0074] For the introduction of the first neural network, the first channel sample information, the second channel sample information, and other operations, please refer to the third aspect, which will not be elaborated here.

[0075] In a seventh aspect, a device is provided. The device can be a terminal device, a device in the terminal device, or a device that can be used in matching with the terminal device.

[0076] In a possible design, the device may include modules corresponding one by one to the methods / operations / steps / actions described in the second aspect. The module can be a hardware circuit, software, or a combination of hardware circuit and software. In one design, the device may include a processing module and a transceiver module. Exemplarily,

[0077] The processing module is configured to perform channel estimation based on the first reference signal received from the first device to determine the first channel sample information;

[0078] The transceiver module is configured to send the first channel sample information to the first device;

[0079] The transceiver module is further configured to receive the information of the third neural network from the first device, where the third neural network is used for the second device to transmit target information to the first device.

[0080] For the introduction of the first reference signal, the first channel sample information, the third neural network, and other operations, please refer to the second aspect, which will not be elaborated here.

[0081] In a possible design, the device may include modules corresponding one by one to the methods / operations / steps / actions described in the fourth aspect. The module can be a hardware circuit, software, or a combination of hardware circuit and software. In one design, the device may include a processing module and a transceiver module. Exemplarily,

[0082] The processing module is configured to perform channel estimation based on the first reference signal received from the first device to determine the first channel sample information;

[0083] The processing module is further configured to determine the first neural network, where the first neural network is trained based on the first channel sample information;

[0084] The transceiver module is configured to send the information of the first neural network to the first device.

[0085] For the introduction of the first reference signal, the first channel sample information, the first neural network, and other operations, please refer to the fourth aspect, which will not be elaborated here.

[0086] In an eighth aspect, a device is provided. The device can be an AI node, a device in the AI node, or a device that can be used in matching with the AI node.

[0087] In a possible design, the apparatus may include modules corresponding one by one to the methods / operations / steps / actions described in the fifth aspect. The module may be a hardware circuit, software, or a combination of hardware circuit and software. In one design, the apparatus may include a processing module and a transceiver module. Exemplarily,

[0088] The transceiver module is used to receive first channel sample information from the first device;

[0089] The processing module is used to train a first neural network according to the first channel sample information, and the first neural network is used to infer second channel sample information.

[0090] In a possible design, the processing module is further used to train a second neural network and / or a third neural network according to the second channel sample information; the transceiver module is further used to send information of the second neural network and / or the third neural network to the first device, and the second neural network and / or the third neural network are used for the first device to transmit target information to the second device.

[0091] In a ninth aspect, an embodiment of the present application provides an apparatus.

[0092] In a possible design, the apparatus includes a processor for implementing the method described in the first aspect above. The apparatus may further include a memory for storing instructions and data. The memory is coupled to the processor, and when the processor executes the instructions stored in the memory, the method described in the first aspect above can be implemented. The apparatus may further include a communication interface for communicating the apparatus with other devices. Exemplarily, the communication interface may be a transceiver, a circuit, a bus, a module, or other types of communication interfaces. In a possible design, the apparatus includes:

[0093] A memory for storing program instructions;

[0094] A processor for receiving first channel sample information from a second device by using the communication interface;

[0095] The processor is further used to determine a first neural network, which is trained according to the first channel sample information and is used to infer second channel sample information.

[0096] For the introduction of the first neural network, the first channel sample information, the second channel sample information, and other operations, please refer to the first aspect, which will not be elaborated here.

[0097] In a possible design, the device includes a processor for implementing the method described in the third aspect above. The device may further include a memory for storing instructions and data. The memory is coupled to the processor, and when the processor executes the instructions stored in the memory, the method described in the third aspect above can be implemented. The device may further include a communication interface for communicating the device with other devices. Exemplarily, the communication interface may be a transceiver, a circuit, a bus, a module, or other types of communication interfaces. In a possible design, the device includes:

[0098] A memory for storing program instructions;

[0099] A processor for using the communication interface to send a first reference signal to a second device; and receiving information of a first neural network from the second device; the first neural network is used to infer second channel sample information.

[0100] For the introduction of the first neural network, the first channel sample information, the second channel sample information, and other operations, please refer to the third aspect, which will not be elaborated here.

[0101] In a tenth aspect, an embodiment of the present application provides a device.

[0102] In a possible design, the device includes a processor for implementing the method described in the second aspect above. The device may further include a memory for storing instructions and data. The memory is coupled to the processor, and when the processor executes the instructions stored in the memory, the method described in the second aspect above can be implemented. The device may further include a communication interface for communicating the device with other devices. Exemplarily, the communication interface may be a transceiver, a circuit, a bus, a module, or other types of communication interfaces. In a possible design, the device includes:

[0103] A memory for storing program instructions;

[0104] A processor for using the communication interface to perform channel estimation according to a first reference signal received from a first device, determining first channel sample information, sending the first channel sample information to the first device, and receiving information of a third neural network from the first device, the third neural network being used for the second device and the first device to transmit target information.

[0105] For the introduction of the first reference signal, the first channel sample information, the third neural network, and other operations, please refer to the second aspect, which will not be elaborated here.

[0106] In a possible design, the device includes a processor for implementing the method described in the fourth aspect above. The device may further include a memory for storing instructions and data. The memory is coupled to the processor, and when the processor executes the instructions stored in the memory, the method described in the fourth aspect above can be implemented. The device may further include a communication interface for communicating the device with other devices. Exemplarily, the communication interface may be a transceiver, a circuit, a bus, a module, or other types of communication interfaces. In a possible design, the device includes:

[0107] A memory for storing program instructions;

[0108] A processor for performing channel estimation based on a first reference signal received from a first device using the communication interface, determining first channel sample information, determining a first neural network, and sending information about the first neural network to the first device, where the first neural network is trained based on the first channel sample information;

[0109] For the introduction of the first reference signal, the first channel sample information, the first neural network, and other operations, please refer to the fourth aspect, which will not be elaborated here.

[0110] In an eleventh aspect, an embodiment of the present application provides a device. The device includes a processor for implementing the method described in the fifth aspect above. The device may further include a memory for storing instructions and data. The memory is coupled to the processor, and when the processor executes the instructions stored in the memory, the method described in the fifth aspect above can be implemented. The device may further include a communication interface for communicating the device with other devices. Exemplarily, the communication interface may be a transceiver, a circuit, a bus, a module, or other types of communication interfaces. In a possible design, the device includes:

[0111] A memory for storing program instructions;

[0112] A processor for receiving first channel sample information from a first device using the communication interface and training a first neural network based on the first channel sample information, where the first neural network is used to infer second channel sample information.

[0113] In a possible design, the processor is further configured to train a second neural network and / or a third neural network based on the second channel sample information; the processor further sends information about the second neural network and / or the third neural network to the first device using the communication interface, where the second neural network and / or the third neural network are used for the first device to transmit target information to a second device.

[0114] In a twelfth aspect, an embodiment of the present application further provides a computer-readable storage medium, including instructions that, when running on a computer, cause the computer to execute any one of the methods in the first aspect to the fifth aspect.

[0115] In a thirteenth aspect, an embodiment of the present application further provides a computer program product, including instructions that, when running on a computer, cause the computer to execute any one of the methods in the first aspect to the fifth aspect.

[0116] In a fourteenth aspect, an embodiment of the present application provides a chip system, which includes a processor and may further include a memory, and is used to implement any one of the methods in the first aspect to the fifth aspect. The chip system may be composed of chips or may include chips and other discrete devices.

[0117] In a fifteenth aspect, an embodiment of the present application further provides a communication system, which includes:

[0118] the device in the sixth aspect and the device in the seventh aspect;

[0119] the device in the sixth aspect, the device in the seventh aspect, and the device in the eighth aspect;

[0120] the device in the ninth aspect and the device in the tenth aspect;

[0121] the device in the ninth aspect, the device in the tenth aspect, and the device in the eleventh aspect. Description of the Drawings

[0122] Figure 1 It is a schematic diagram of the network architecture provided by an embodiment of the present application;

[0123] Figure 2 It is a schematic diagram of the hardware structure of the communication device provided by an embodiment of the present application;

[0124] Figure 3 It is a schematic diagram of the neuron structure provided by an embodiment of the present application;

[0125] Figure 4 It is a schematic diagram of the layer relationship of the neural network provided by an embodiment of the present application;

[0126] Figure 5 It is a schematic diagram of the convolutional neural network CNN provided by an embodiment of the present application;

[0127] Figure 6 It is a schematic diagram of the feedforward neural network RNN provided by an embodiment of the present application;

[0128] Figure 7 It is a schematic diagram of the generative adversarial network GAN provided by an embodiment of the present application;

[0129] Figure 8Schematic diagram of the variational autoencoder provided by the embodiment of the present application;

[0130] Figure 9 Schematic diagram of the neural network architecture for transceiver joint optimization of constellation modulation / demodulation provided by the embodiment of the present application;

[0131] Figures 10 - 12 Schematic flowchart of the training method of the neural network provided by the embodiment of the present application;

[0132] Figure 13 Schematic diagram of the structure of a generator network of the first neural network provided by the embodiment of the present application;

[0133] Figure 14 Schematic diagram of the structure of a discriminator network of the first neural network provided by the embodiment of the present application;

[0134] Figure 15a Schematic diagram of the structure of a generator network of the first neural network provided by the embodiment of the present application;

[0135] Figure 15b Schematic diagram of the structure of a discriminator network of the first neural network provided by the embodiment of the present application;

[0136] Figure 16a and Figure 16b Schematic diagram of the network structure provided by the embodiment of the present application;

[0137] Figure 17 Schematic diagram of a communication device provided by the embodiment of the present application. Detailed implementation manners

[0138] In a wireless communication system, including communication devices, the communication devices can perform wireless communication by using wireless resources. Among them, the communication devices can include access network devices and terminal devices, and the access network devices can also be referred to as access side devices. The wireless resources can include link resources and / or air interface resources. The air interface resources can include at least one of time domain resources, frequency domain resources, code resources, and space resources. In the embodiment of the present application, at least one (piece) can also be described as one (piece) or more (pieces)), and multiple (pieces) can be two (pieces), three (pieces), four (pieces) or more (pieces), which is not limited in the present application.

[0139] In the embodiments of the present application, " / " can indicate that the objects associated before and after are in an "or" relationship. For example, A / B can represent A or B; "and / or" can be used to describe three relationships of the associated objects. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. To facilitate the description of the technical solutions in the embodiments of the present application, in the embodiments of the present application, terms such as "first" and "second" can be used to distinguish technical features with the same or similar functions. These terms such as "first" and "second" do not limit the quantity and execution order, and these terms such as "first" and "second" do not necessarily limit being different. In the embodiments of the present application, words such as "exemplary" or "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as "exemplary" or "for example" should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Using words such as "exemplary" or "for example" aims to present relevant concepts in a specific way for easy understanding.

[0140] As Figure 1 shown, it is a schematic diagram of the network architecture applicable to the embodiments of the present application. The communication system in the embodiments of the present application can be a system including an access network device (such as Figure 1 the base station shown) and a terminal device, or can be a system including two or more terminal devices. In this communication system, the access network device can send configuration information to the terminal device, and the terminal device performs corresponding configurations according to the configuration information. The access network device can send downlink data to the terminal device, and / or the terminal device can send uplink data to the access network device. In a communication system including two or more terminal devices (such as a vehicle-to-everything network), terminal device 1 can send configuration information to terminal device 2, terminal device 2 performs corresponding configurations according to the configuration information, terminal device 1 can send data to terminal device 2, and terminal device 2 can also send data to terminal device 1. Optionally, Figure 1 in the communication system shown, the access network device can implement one or more of the following artificial intelligence (AI) functions: model training and inference. Optionally, Figure 1In the communication system shown, the network side may include nodes independent of the access network device for implementing one or more of the following AI functions: model training and inference. This node may be referred to as an AI node, a model training node, an inference node, a wireless intelligent controller, or other names, without limitation. Exemplarily, the model training function and the inference function may be implemented by the access network device; or, the model training function and the inference function may be implemented by the AI node; or, the model training function may be implemented by the AI node, and the information of the model may be sent to the access network device, and the inference function may be implemented by the access network device. Optionally, if the AI node implements the inference function, the AI node may send the inference result to the access network device for the access network device to use, and / or the AI node may send the inference result to the terminal device through the access network device for the terminal device to use; if the access network device implements the inference function, the access network device may use the inference result, or send the inference result to the terminal device for the terminal network device to use. If the AI node is used to implement the model training function and the inference function, the AI node may be separated into two nodes, one of which implements the model training function and the other implements the inference function.

[0141] The embodiments of this application do not limit the specific number of network elements in the involved communication system.

[0142] The terminal device involved in the embodiments of this application can also be referred to as a terminal or an access terminal, and can be a device with wireless transceiver functions. The terminal device can communicate with one or more core networks (CNs) via an access network device. The terminal device can be a subscriber unit, a subscriber station, a mobile station, a mobile unit, a remote station, a remote terminal, a mobile device, a user terminal, a user equipment (UE), a user agent, or a user device, etc. The terminal device can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; it can also be deployed on water (such as a ship, etc.); it can also be deployed in the air (such as an airplane, a balloon, a satellite, etc.). The terminal device can be a cellular phone, a cordless phone, a session initiation protocol (SIP) phone, a smart phone, a mobile phone, a wireless local loop (WLL) station, a personal digital assistant (PDA). The terminal device can also be a handheld device, a computing device, or other devices with wireless communication functions, a vehicle-mounted device, a wearable device, a drone device, a terminal in the Internet of Things or the Internet of Vehicles, a terminal in the fifth generation (5G) mobile communication network, a relay user equipment, or a terminal in a future evolved mobile communication network, etc. Among them, the relay user equipment can be, for example, a 5G residential gateway (RG). For another example, the terminal device can be a virtual reality (VR) terminal, an augmented reality (AR) terminal, a wireless terminal in industrial control, a wireless terminal in self-driving, a wireless terminal in remote medical, a wireless terminal in smart grid, a wireless terminal in transportation safety, a wireless terminal in smart city, or a wireless terminal in smart home, etc. The embodiments of this application do not limit this. In the embodiments of this application, the device for implementing the functions of the terminal can be the terminal; it can also be a device capable of supporting the terminal to implement the functions, such as a chip system, and this device can be installed in the terminal or used in matching with the terminal. In the embodiments of this application, the chip system can be composed of chips, or can include chips and other discrete devices.

[0143] An access network device can be regarded as a sub-network of an operator's network and is an implementation system between service nodes and terminal devices in the operator's network. When a terminal device wants to access the operator's network, it can first pass through the access network device and then connect to the service nodes of the operator's network through the access network device. The access network device in the embodiments of the present application is a device located in a (radio) access network ((R)AN) that can provide wireless communication functions for terminal devices. The access network device includes a base station, such as but not limited to: the next generation node B (gNB) in a 5G system, the evolved node B (eNB) in a long term evolution (LTE) system, a radio network controller (RNC), a node B (NB), a base station controller (BSC), a base transceiver station (BTS), a home base station (e.g., home evolved nodeB, or home node B, HNB), a base band unit (BBU), a transmitting and receiving point (TRP), a transmitting point (TP), a pico base station device, a mobile switching center, or an access network device in a future network, etc. In systems using different radio access technologies, the names of the devices with the functions of access network devices may be different. In the embodiments of the present application, the device for implementing the functions of the access network device can be the access network device; it can also be a device that can support the access network device to implement the functions, such as a chip system, and this device can be installed in the access network device or used in matching with the access network device.

[0144] The technical solutions provided in the embodiments of the present application can be applied to various communication systems, such as: LTE systems, 5G systems, wireless-fidelity (WiFi) systems, future sixth-generation mobile communication systems, or systems integrating multiple communication systems, etc., which are not limited in the embodiments of the present application. Among them, 5G can also be referred to as new radio (NR).

[0145] The technical solutions provided by the embodiments of this application can be applied to various communication scenarios. For example, they can be applied to one or more of the following communication scenarios: enhanced mobile broadband (eMBB) communication, ultra-reliable low-latency communication (URLLC), machine-type communication (MTC), massive machine-type communication (mMTC), device-to-device (D2D) communication, vehicle-to-everything (V2X) communication, vehicle-to-vehicle (V2V) communication, and Internet of Things (IoT), etc. In the embodiments of this application, the term "communication" can also be described as "transmission", "information transmission", "data transmission", or "signal transmission", etc. Transmission can include sending and / or receiving. In the embodiments of this application, the technical solutions are described by taking the communication between an access network device and a terminal device as an example. Those skilled in the art can also use this technical solution for the communication between other scheduling entities and subordinate entities, such as the communication between a macro base station and a micro base station, and / or the communication between terminal device 1 and terminal device 2, etc.

[0146] In addition, the network architecture and service scenarios described in this application are for more clearly explaining the technical solutions of this application, and do not constitute a limitation on the technical solutions provided by this application. Those skilled in the art know that with the evolution of the network architecture and the emergence of new service scenarios, the technical solutions provided by this application are equally applicable to similar technical problems.

[0147] Figure 2This is a schematic diagram of the hardware structure of the communication device in the embodiments of the present application. The communication device may be a possible implementation of the AI node, access network device, or terminal device in the embodiments of the present application. The communication device may be an AI node, a device in the AI node, or a device that can be used in combination with the AI node. The communication device may be an access network device, a device in the access network device, or a device that can be used in combination with the access network device. The communication device may be a terminal device, a device in the terminal device, or a device that can be used in combination with the terminal device. Among them, the device may be a chip system. In the embodiments of the present application, the chip system may be composed of chips or may include chips and other discrete devices. The connections in the embodiments of the present application are indirect couplings or communication connections between devices, units, or modules, which may be electrical, mechanical, or other forms for information interaction between devices, units, or modules.

[0148] As Figure 2 shown, the communication device includes at least one processor 204 for implementing the technical solutions provided in the embodiments of the present application. Optionally, the communication device may further include a memory 203. The memory 203 is used to store instructions 2031 and / or data 2032. The memory 203 is connected to the processor 204. The processor 204 and the memory 203 may operate cooperatively. The processor 204 may execute the instructions stored in the memory 203 to implement the technical solutions provided in the embodiments of the present application. The communication device may further include a transceiver 202 for receiving and / or transmitting signals. Optionally, the communication device may further include one or more of the following: an antenna 206, an I / O (Input / Output) interface 210, and a bus 212. The transceiver 202 further includes a transmitter 2021 and a receiver 2022. The processor 204, the transceiver 202, the memory 203, and the I / O interface 210 are communicatively connected to each other through the bus 212, and the antenna 206 is connected to the transceiver 202. The bus 212 may include an address bus, a data bus, and / or a control bus, etc. Figure 2 In [FIGURE] only a thick line is used to represent the bus 212, but it does not mean that there is only one bus or one type of bus for the bus 212.

[0149] The processor 204 may be a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field-programmable gate array, and / or other programmable logic devices, such as discrete gate or transistor logic devices and / or discrete hardware components. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor 204 may implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present application. The steps of the methods disclosed in combination with the embodiments of the present application may be embodied as being executed by a hardware processor, or executed by a combination of hardware and software modules in the processor. Exemplarily, the processor 204 may be a central processing unit (CPU), or a dedicated processor, such as, but not limited to, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), and / or a field-programmable gate array (FPGA), etc. The processor 204 may also be a neural processing unit (NPU). In addition, the processor 204 may also be a combination of multiple processors. In the technical solution provided by the embodiments of the present application, the processor 204 may be used to execute the relevant steps in the subsequent method embodiments. The processor 204 may be a processor specifically designed to execute the above steps and / or operations, or a processor that executes the above steps and / or operations by reading and executing the instructions 2031 stored in the memory 203. The processor 204 may need to use the data 2032 during the execution of the above steps and / or operations.

[0150] The transceiver 202 includes a transmitter 2021 and a receiver 2022. In an alternative implementation, the transmitter 2021 is used to transmit signals through at least one of the antennas 206. The receiver 2022 is used to receive a second reference signal through at least one of the antennas 206.

[0151] In the embodiments of the present application, the transceiver 202 is used to support the communication device to execute the receiving function and the sending function. Regarding the processor with processing functions as the processor 204. The receiver 2022 may also be referred to as an input port, a receiving circuit, a receiving bus, or other devices that implement the receiving function, etc. The transmitter 2021 may be referred to as a transmitting port, a transmitting circuit, a transmitting bus, or other devices that implement the transmitting function, etc. The transceiver 202 may also be referred to as a communication interface.

[0152] The processor 204 can be used to execute the instructions stored in the memory 203, for example, to control the transceiver 202 to receive messages and / or send messages, so as to complete the functions of the communication device in the method embodiments of this application. As an implementation manner, the functions of the transceiver 202 can be considered to be implemented by a transceiver circuit or a dedicated transceiver chip. In the embodiments of this application, the transceiver 202 receiving a message can be understood as the transceiver 202 inputting a message, and the transceiver 202 sending a message can be understood as the transceiver 202 outputting a message.

[0153] The memory 203 can be a non-volatile memory, such as a hard disk drive (HDD) or a solid-state drive (SSD), etc., or can be a volatile memory, such as a random-access memory (RAM). A memory is any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory in the embodiments of this application can also be a circuit or any other device capable of implementing a storage function, for storing program instructions and / or data. The memory 203 is specifically used to store instructions 2031 and data 2032. The processor 204 can execute the steps and / or operations in the method embodiments of this application by reading and executing the instructions 2031 stored in the memory 203, and may need to use the data 2032 during the execution of the operations and / or steps in the method embodiments of this application.

[0154] Optionally, the communication device may further include an I / O interface 210, which is used to receive instructions and / or data from a peripheral device, and output instructions and / or data to the peripheral device.

[0155] Next, some concepts related to the embodiments of this application are introduced:

[0156] Machine learning (ML) has attracted extensive attention in the academic and industrial communities in recent years. Due to the huge advantages of machine learning in the face of structured information and massive data, many researchers in the field of communication have also turned their attention to machine learning. Communication technologies based on machine learning have great potential in signal classification, channel estimation, and / or performance optimization. Most communication systems are designed block by block, which means that these communication systems are composed of multiple modules. For such a communication architecture based on modular design, many technologies can be developed to optimize the performance of each module. However, the best performance of each module does not necessarily mean the best performance of the entire communication system. End-to-end optimization (i.e., optimizing the entire communication system) may be better than optimizing a single model. Machine learning provides an advanced and powerful tool to maximize the end-to-end performance as much as possible. In a wireless communication system, in complex and large-scale communication scenarios, the channel conditions change rapidly. Many traditional communication models, such as the large-scale multiple input multiple output (MIMO) model, rely heavily on channel state information, and the performance of these models will deteriorate under non-linear time-varying channels. Therefore, accurately obtaining the channel state information (CSI) of the time-varying channel is crucial for system performance. By using machine learning techniques, it is possible to enable the communication system to learn the mutated channel model and timely feedback the channel state.

[0157] Based on the above considerations, using machine learning techniques in wireless communication can meet the new requirements in future wireless communication scenarios.

[0158] Machine learning is an important technical approach to realizing artificial intelligence. Machine learning can include supervised learning, unsupervised learning, and reinforcement learning.

[0159] Supervised learning is based on the collected sample values and sample labels, uses machine learning algorithms to learn the mapping relationship from sample values to sample labels, and uses a machine learning model to represent the learned mapping relationship. Among them, the sample label can also be simply referred to as the label. The process of training a machine learning model is the process of learning this mapping relationship. For example, in signal detection, the received signal with noise is the sample, and the true constellation point corresponding to this signal is the label. Machine learning expects to learn the mapping relationship between the sample value and the label through training, that is, to make the machine learning model learn a signal detector. During training, the model parameters are optimized by calculating the error between the predicted value of the model and the true label. After the mapping relationship learning is completed, the learned mapping can be used to predict the sample labels of new samples. The mapping relationship learned by supervised learning can include linear mapping or non-linear mapping. According to the type of label, the learning tasks can be divided into classification tasks and regression tasks.

[0160] Unsupervised learning relies on the collected sample values and uses algorithms to automatically discover the internal patterns of the samples. In unsupervised learning, there is a type of algorithm that uses the samples themselves as the supervision signal, that is, the model learns the mapping relationship from samples to samples, which is also called self-supervised learning. During training, the model parameters are optimized by calculating the error between the predicted value of the model and the samples themselves. For example, self-supervised learning can be used in applications such as signal compression and decompression recovery. Common algorithms include autoencoders and generative adversarial networks, etc.

[0161] Reinforcement learning is different from supervised learning. It is a type of algorithm that learns strategies to solve problems by interacting with the environment. Different from supervised and unsupervised learning, there is no clear "correct" action label data in reinforcement learning problems. The algorithm needs to interact with the environment to obtain the reward signal feedback from the environment, and then adjust the decision-making actions to obtain a larger numerical value of the reward signal. For example, in downlink power control, the reinforcement learning model adjusts the downlink transmission power of each user according to the total system throughput rate fed back by the wireless network, and then expects to obtain a higher system throughput rate. The goal of reinforcement learning is also to learn the mapping relationship between the environmental state and the optimal decision-making action. However, because the label of the "correct action" cannot be obtained in advance, the network cannot be optimized by calculating the error between the action and the "correct action". The training of reinforcement learning is achieved through iterative interaction with the environment.

[0162] Deep neural network (DNN) is a specific implementation form of machine learning. According to the universal approximation theorem, DNN can theoretically approximate any continuous function, enabling DNN to have the ability to learn any mapping. Traditional communication systems need to rely on rich expert knowledge to design communication modules, while deep learning communication systems based on DNN can automatically discover implicit pattern structures from large datasets, establish the mapping relationship between data, and obtain performance superior to traditional modeling methods.

[0163] The idea of DNN comes from the neuron structure of the brain tissue. Each neuron performs a weighted summation operation on its input values, and passes the weighted summation result through an activation function to generate an output. As Figure 3 shown, Figure 3 is the schematic diagram of the neuron structure. Assume that the input of the neuron is x = [x 0 , x 1 , … x n , the weights corresponding to the inputs are w = [w 0 , w 1 , … w n , the bias of the weighted summation is b, and the form of the activation function can be diversified. As an example, if the activation function of a neuron is: y = f(x) = max{0, x}, then the output executed by a neuron is: wherein, w i x i represents the product of w i and x i , the data type of b is integer or floating point number, and the data type of w i is integer or floating point number. A DNN generally has a multi-layer structure, and each layer of the DNN can contain one or more neurons. After the numerical values received by the input layer of the DNN are processed by the neurons, they are passed to the intermediate hidden layer. Similarly, the hidden layer then passes the calculation results to the adjacent next hidden layer or the adjacent output layer to generate the final output of the DNN. As Figure 4 shown Figure 4 is a schematic diagram of the layer relationship of the neural network.

[0164] A DNN generally has one or more hidden layers, and the hidden layers can affect the ability to extract information and fit functions. Increasing the number of hidden layers of the DNN or expanding the number of neurons in each layer can improve the function fitting ability of the DNN. The parameters of each neuron include weights, biases, and activation functions, and the set of parameters of all neurons in the DNN is called DNN parameters (or neural network parameters). The weights and biases of the neurons can be optimized through the training process, so that the DNN has the ability to extract data features and express mapping relationships. The DNN generally uses supervised learning or unsupervised learning strategies to optimize the neural network parameters.

[0165] According to the construction method of the network, the DNN can include a feedforward neural network (FNN), a convolutional neural network (CNN), and a recurrent neural network (RNN).

[0166] Figure 4 Shown is an FNN, whose characteristic is that the neurons between adjacent layers are completely connected in pairs, which makes the FNN usually require a large amount of storage space and result in a high computational complexity.

[0167] As Figure 5 shown Figure 5It is a schematic diagram of a CNN. A CNN is a neural network used to process data with a similar grid structure. For example, time series data (discretely sampled on the time axis) and image data (two-dimensionally discretely sampled) can both be considered data with a similar grid structure. Instead of using all the input information for calculation at once, a CNN uses a window (such as a window of a fixed size) to intercept part of the information for convolution operations, which greatly reduces the computational amount of the neural network parameters. Additionally, according to the different types of information intercepted by the window (such as people and objects in the same picture being different types of information), different convolution kernels can be used for each window, enabling the CNN to better extract the features of the input data. Among them, the convolutional layer is used for feature extraction to obtain a feature map. The pooling layer is used to compress the input feature map to make the feature map smaller and simplify the network computational complexity. The fully connected layer is used to map the learned "distributed feature representation" to the sample label space. Exemplarily, Figure 5 in which, the probability that the image is the sun is 0.7, the probability that it is the moon is 0.1, the probability that it is a car is 0.05, and the probability that it is a house is 0.02.

[0168] such as Figure 6 shown Figure 6 It is a schematic diagram of an RNN. An RNN is a type of DNN that utilizes feedback time series information, and its input includes the new input value at the current moment and its own output value at the previous moment. RNNs are suitable for obtaining sequential features that are temporally correlated and are particularly applicable to applications such as speech recognition and channel coding. Referring to Figure 6 , the input of a neuron at multiple moments generates multiple outputs. For example, at the 0th moment, the inputs are x 0 and s 0 , and the outputs are y 0 and s 1 , at the 1st moment, the inputs are x 1 and s 1 , and the outputs are y 1 and s 2 , ……, at the tth moment, the inputs are x t and s t , and the outputs are y t and s t+1 .

[0169] A generative neural network is a special type of deep learning neural network. Different from general neural networks that mainly perform classification and prediction tasks, a generative neural network can learn the probability distribution function followed by a set of training samples, so it can be used to model random variables and establish the conditional probability distribution between variables. Common generative neural networks include generative adversarial networks (GANs) and variational autoencoders (VAEs).

[0170] Please refer to Figure 7 , Figure 7 which is a schematic diagram of the generative adversarial network GAN. The GAN network can be divided into two parts according to its functions: a generator network (simply called the generator) and a discriminator network (simply called the discriminator). Specifically, the generator network processes the input random noise and outputs generated samples. The discriminator network compares the generated samples output by the generator network with the training samples in the training set to determine whether the generated samples and the training samples approximately follow similar probability distributions and outputs a judgment of true or false. For example, taking the case where the generated samples and the training samples follow a normal distribution, when the mean of the generated samples is the same as the mean of the training samples and the difference between the variance of the generated samples and the variance of the training samples is less than a threshold (such as 0.01), it is considered that the generated samples and the training samples follow similar probability distributions. There is a game relationship between the generator network and the discriminator network: the generator network hopes to generate samples that follow the distribution of the training set as much as possible, and the discriminator network hopes to distinguish the differences between the generated samples and the training set as much as possible. By jointly training the generator network and the discriminator network, the two can reach an equilibrium state, that is, the probability distribution followed by the generated samples output by the generator network is similar to the probability distribution followed by the training samples, and the discriminator network believes that the generated samples and the training set follow a similar distribution. Optionally, this similar distribution can be called the same distribution.

[0171] Please refer to Figure 8 , Figure 8 which is a schematic diagram of the variational autoencoder. The VAE can be divided into three parts according to its functions: an encoder network (simply called the encoder), a decoder network (simply called the decoder), and a discriminator network (simply called the discriminator). The encoder network compresses the samples in the input training set into intermediate variables, and the decoder network attempts to restore the intermediate variables to the samples in the training set. At the same time, certain constraints can be imposed on the form of the intermediate variables. Similar to the discriminant network in GAN, a discriminant network can also be used in VAE to determine whether the intermediate variables follow the distribution of random noise. By jointly training the encoder network, the decoder network, and the discriminator network, the decoder network can use the input random noise to generate generated samples that conform to the distribution of the training set.

[0172] The above-mentioned FNN, CNN, RNN, GAN, and VAE are neural network structures, and these network structures are all constructed based on neurons.

[0173] Benefiting from the advantages of machine learning in modeling and extracting information features, communication schemes based on machine learning can be designed, which can achieve good performance. This includes but is not limited to CSI compression feedback, adaptive constellation point design, and / or robust precoding, etc. These schemes optimize the transmission performance or reduce the processing complexity by replacing the sending or receiving module in the original communication system with a neural network model. To support different application scenarios, different neural network model information can be predefined or configured, so that the neural network model can adapt to the requirements of different scenarios.

[0174] In the embodiments of this application, in a communication system based on a neural network, some communication modules of the access network device and / or the terminal device can adopt a neural network model. As Figure 9 shown, Figure 9 is a schematic diagram of a neural network architecture for transceiver joint optimization of constellation modulation / demodulation. Among them, the constellation mapping neural network at the sending end (also called the transmitting end) and the constellation demapping neural network at the receiving end (also called the receiving end) both adopt neural network models. The constellation mapping neural network at the transmitting end maps the bit stream into constellation symbols, and the constellation demapping neural network demaps (or demodulates) the received constellation symbols into the log-likelihood ratio of bit information. By collecting channel data to train the neural network, better end-to-end communication performance can be achieved. In a communication system, the transmitting end performs a series of processes on the bit stream to be sent, which may include one or more of the following: channel coding, constellation symbol mapping modulation, orthogonal frequency division multiplexing (OFDM) modulation, layer mapping, precoding, and upconversion, etc. For the sake of convenience of expression, Figure 9 only constellation mapping symbol modulation and OFDM modulation are listed.

[0175] It should be noted that the neural network involved in the embodiments of this application is not limited to any specific application scenario, but can be applied to any communication scenario, such as CSI compression feedback, adaptive constellation point design, and / or robust precoding, etc.

[0176] The above neural network needs to be adaptively trained to ensure communication performance. For example, for each different set of wireless system parameters (including one or more of the wireless channel type, bandwidth, number of receiving antennas, number of transmitting antennas, modulation order, number of paired users, channel coding method, and coding rate), a corresponding set of neural network model information (including neural network structure information and neural network parameters) is defined. Taking the adaptive modulation constellation design based on artificial intelligence (AI) as an example, when the number of transmitting antennas is 1 and the number of receiving antennas is 2, a set of neural network model information is required to generate the corresponding modulation constellation; when the number of transmitting antennas is 1 and the number of receiving antennas is 4, another set of neural network model information is required to generate the corresponding modulation constellation. Similarly, for different wireless channel types, bandwidths, modulation orders, numbers of paired users, channel coding methods, and / or coding rates, the corresponding neural network model information may be different.

[0177] To enable the (transmitting and receiving end) neural network to achieve good performance in actual channel conditions, actual channel sample information needs to be used for joint training when training the neural network. Generally speaking, the larger the number of channel sample information used in training, the better the training effect. The sources of this channel sample information include: 1. Obtained by the receiving end (such as the UE) according to actual measurements; 2. Generated using a mathematical (channel) model.

[0178] Specifically as follows: 1. The receiving end of the signal (such as a reference signal, or a synchronization signal, etc.) (such as the UE) measures the real channel to obtain channel sample information. This channel sample information can more accurately reflect the channel environment. If the network element used to train the neural network using the channel sample information is the transmitting end of the signal, after the receiving end measures the channel sample information, it feeds back the channel sample information to the transmitting end (such as the base station). The transmitting end trains the neural network according to this channel sample information. In order to improve the training effect and make the performance of the trained neural network better, the receiving end of the signal needs to feed back a large amount of channel sample information to the transmitting end of the signal. However, feeding back a large amount of channel sample information will occupy a large amount of air interface resources and affect the data transmission efficiency between the transmitting and receiving ends.

[0179] 2. In a possible implementation, for different channel types, a mathematical expression can be used to model the channel model. That is, a mathematical (channel) model is used to simulate the real channel. For example: in the protocol, mathematical channel models such as tapped delay line (TDL) and clustered delay line (CDL) can be defined. Each channel type can be further divided into multiple subcategories. For example, the TDL and CDL channels each include five subcategories A, B, C, D, and E; each subcategory is further divided into multiple typical channel scenarios according to specific parameters. For example, it is divided into multiple typical scenarios such as 10 nanoseconds (ns), 30 ns, 100 ns, 300 ns, or 1000 ns according to the multipath delay interval. Therefore, when training the transceiver neural network, a mathematical channel model similar to the actual environment can be selected to generate a large number of channel sample information similar to the actual channel, and this channel sample information can be used for training.

[0180] Although using a mathematical channel model can greatly reduce the signaling overhead for obtaining channel samples, there are also disadvantages of mismatch between the mathematical channel model and the actual channel model. For example, the TDL channel model assumes a limited number of reflection paths, and the channel coefficient of each path follows a simple Rayleigh distribution. However, the number of reflection paths in the actual channel varies in different environments, and the Rayleigh distribution cannot accurately describe the distribution of the channel coefficient of each path. In addition, the multipath delay interval also often varies depending on the scenario, and coarsely dividing it into several typical values will inevitably cause modeling errors. Therefore, the mathematical channel modeling method is difficult to accurately describe the actual channel model, thereby affecting the training effect of the neural network. That is, there is a problem of data model mismatch.

[0181] In summary, how to generate channel sample information close to the actual channel scenario under the premise of saving air interface resources has become an urgent problem to be solved currently.

[0182] Next, the technical solution proposed in this application will be introduced in combination with the accompanying drawings. Please refer to Figure 10 , Figure 10 is a schematic flow chart of a method for training a neural network proposed in an embodiment of this application. A method for training a neural network proposed in an embodiment of this application includes:

[0183] Optionally, 1001. The first device sends a first reference signal to the second device.

[0184] In this embodiment, the first device is taken as an access network device and the second device is taken as a terminal device for illustration. It can be understood that the first device can be an access network device or a chip in the access network device, and the second device can be a terminal device or a chip in the terminal device. The first device can also be a terminal device or a chip in the terminal device, and the second device can also be an access network device or a chip in the access network device, which is not limited here.

[0185] The first device sends a first reference signal to the second device. The first reference signal can be a synchronization signal, a synchronization signal and PBCH block (SSB), a demodulation reference signal (DMRS), or a channel state information reference signal (CSI-RS), etc. The first reference signal can also be referred to as the first signal. The first reference signal can also be a newly defined reference signal, which is not limited here. The sequence type of the first reference signal includes, but is not limited to: Zadoff-Chu (ZC) sequence or Gold sequence. Optionally, the DMRS can be the DMRS of the physical downlink control channel (PDCCH) or the DMRS of the physical downlink shared channel (PDSCH).

[0186] Optionally, before step 1001, when the second device needs to use a neural network (the neural network used by the second device is called the third neural network), the second device sends capability information to the first device. The capability information is used to indicate one or more of the following information of the second device:

[0187] 1) Whether it supports using a neural network to replace or implement the functions of the communication module;

[0188] The communication module includes, but is not limited to: OFDM modulation module, OFDM demodulation module, constellation mapping module, constellation demapping module, channel coding module, channel decoding module, precoding module, equalization module, interleaving module, and / or deinterleaving module.

[0189] 2) Whether it supports the network type of the third neural network, or indicates the supported network type of the third neural network;

[0190] Among them, the network type of the third neural network includes one or more of the following: fully connected neural network, radial basis neural network, convolutional neural network, recurrent neural network, Hopfield neural network, restricted Boltzmann machine, or deep belief network, etc. The third neural network can be any of the above neural networks, and the third neural network can also be a combination of the above multiple neural networks, which is not limited here.

[0191] 3) Whether it supports receiving information of the third neural network through signaling;

[0192] 4) Stored pre-configured third neural network;

[0193] For example: If there is already a predefined or pre-configured third neural network in the second device, the identifier or index of the third neural network can be carried in the capability information.

[0194] Optionally, the identifier or index of other predefined or pre-configured neural networks already existing in the second device can also be carried in the capability information.

[0195] 5) Memory space available for storing the third neural network;

[0196] 6) Computing power information available for running the neural network;

[0197] This computing power information refers to the computing ability information for running the neural network, such as including information such as the operation speed of the processor and / or the amount of data that the processor can process.

[0198] Optionally, 1002. The second device performs channel estimation based on the received first reference signal from the first device to determine the first channel sample information.

[0199] In this embodiment, after receiving the first reference signal, the second device performs channel estimation on the first reference signal to determine the first channel sample information.

[0200] Specifically, for ease of description, the signal received at the second device after the first reference signal is transmitted through the channel is called the second reference signal. Let the first reference signal be x and the second reference signal be y, then the channel for propagating the first reference signal can be understood as a function with a transition probability P(y|x).

[0201] The second device can pre-configure the information of the first reference signal for the first device through signaling, or the information of the first reference signal can be agreed upon by the protocol, without limitation. In the embodiments of the present application, the type of signaling is not limited. For example, it can be a broadcast message, system information, radio resource control (RRC) signaling, media access control (MAC) control element (CE), or downlink control information (DCI).

[0202] Since the second device already knows the information of the first reference signal x sent by the first device in advance, when the second device receives the second reference signal y, the second device performs channel estimation based on the second reference signal y and the transmitted first reference signal x to determine the first channel sample information. At this time, the channel experienced by the first reference signal from being transmitted to being received can be estimated, such as the amplitude change and phase rotation experienced. The first channel sample information includes, but is not limited to, the second reference signal and / or channel state information (CSI).

[0203] The channel estimation algorithms used by the second device for channel estimation include, but are not limited to: least square (LS) or linear minimum mean square error (LMMSE).

[0204] 1003. The second device sends the first channel sample information to the first device.

[0205] In this embodiment, the second device sends the first channel sample information to the first device.

[0206] 1004. The first device trains a first neural network based on the first channel sample information.

[0207] In this embodiment, the first device trains a first neural network based on the first channel sample information. The first neural network is used to generate new channel sample information.

[0208] Optionally, the first neural network is a generative neural network. In the embodiments of the present application, taking the first neural network as a GAN as an example for illustration, it can be understood that the first neural network can be other types of neural networks, such as VAE, etc., which are not limited herein. According to the function, the first neural network includes a generator network and a discriminator network. Among them, the generator network is used to generate the second channel sample information, and the discriminator network is used to determine whether the newly generated second channel sample information and the first channel sample information from the second device follow similar probability statistical characteristics. By using the method of machine learning, jointly training the generator network and the discriminator network of the first neural network can make the second channel sample information output by the generator network converge to the probability distribution of the first channel sample information from the second device. The following are examples for illustration respectively.

[0209] Exemplarily, when the first channel sample information is CSI (denote CSI as h), please refer to the schematic structural diagram of the first neural network Figures 13 - 14 . Figure 13 This is the schematic structural diagram of a generator network of the first neural network in the embodiments of the present application. Figure 14This is a schematic diagram of a discriminator network of the first neural network in the embodiments of the present application. Figures 13 - 14 In the first neural network shown, the generator network includes a 5-layer convolutional layer network. The input of the generator network is random noise z, which includes but is not limited to Gaussian white noise. The output of the generator network is the second channel sample information The discriminator network includes a 3-layer convolutional layer network and a 3-layer fully connected layer network. Part of the input of the discriminator network is the output information of the generator network Another part of the input signal of the discriminator network includes the first channel sample information h from the second device. The output c of the discriminator network is a binary variable, representing whether the output second channel sample information of the generator network follows the probability distribution of the first channel sample, that is whether it follows the probability distribution of h.

[0210] Exemplarily, when the first channel sample information is the second reference signal (y) introduced above, please refer to the schematic diagram of the structure of the first neural network Figures 15a - 15b . Figure 15a This is a schematic diagram of a generator network structure of the first neural network in the embodiments of the present application. Figure 15b This is a schematic diagram of a discriminator network structure of the first neural network in the embodiments of the present application.

[0211] Figures 15a - 15b In the first neural network shown, the generator network is composed of a 7-layer convolutional layer network. The input of the generator network includes random noise z and a training sequence x, and the output is a sample The discriminator network includes a 4-layer convolutional layer network and a 4-layer fully connected layer network. The input includes the sample generated by the generator network The sample second reference signal y from the second device, and the corresponding first reference signal x. The output c of the discriminator network is a binary variable, representing whether it follows the probability distribution of P(y|x).

[0212] Assume that the first reference signal sent by the first device is x, and after passing through the channel, the second reference signal received at the second device is y. Then this channel can be understood as a function with a transition probability P(y|x). The first neural network hopes to learn this probability transfer feature of the channel. For each input x, it can generate the output y with the probability of P(y|x), that is, the first neural network simulates the process of signal propagation through the channel.

[0213] Specifically, the first device uses the first channel sample information to train a reference neural network. The neural network obtained after the training is completed is called the first neural network. In the embodiments of the present application, for the convenience of description, the neural network before being trained using the first channel sample information can be called the reference neural network. Exemplarily, the first device uses the first channel sample information including 5,000 samples from the second device to train the reference neural network, so as to obtain the first neural network after training.

[0214] In an alternative implementation, the reference neural network can pre-configure the parameters of some neurons or the structure of the neural network. The reference neural network can also pre-configure the information of other neural networks, which is not limited here. For example: the reference neural network can be a neural network trained in a predefined channel environment for generating channel sample information (such as CSI or the second reference signal, etc.). The first device uses the first channel sample information to further train the reference neural network, and obtains the first neural network after training.

[0215] In another alternative implementation, the reference neural network can be an initialized neural network. For example: the parameters of each neuron in the reference neural network are zero, random values, or other values agreed upon, and / or the structure of the neural network is a random or agreed initial structure, etc., which is not limited here. The first device uses the first channel sample information to start training the reference neural network, so as to obtain the first neural network after training.

[0216] 1005. The first device uses the first neural network to infer the second channel sample information.

[0217] In this embodiment, when the first device obtains the trained first neural network, it uses the first neural network to infer the second channel sample information.

[0218] Exemplarily, the first device inputs random noise into the first neural network, and the second channel sample information can be inferred by the first neural network. Exemplarily, when the first neural network is a GAN, inputting random noise into the generator network of the first neural network can obtain the second channel sample information. Another example, when the first neural network is a VAE, inputting random noise into the decoder network of the first neural network can obtain the second channel sample information.

[0219] The form of the second channel sample information is the same as that of the first channel sample information. For example, when the first channel sample information is CSI(h), the second channel sample information is CSI When the first channel sample information is the second reference signal (y), the second channel sample information is the second reference signal

[0220] Optionally, 1006. The first device trains the second neural network and / or the third neural network according to the second channel sample information.

[0221] Exemplarily, when the first channel sample information is CSI (denoted as h), a schematic structural diagram of the first device training the second neural network and / or the third neural network according to the second channel sample information is shown in Figure 16a . The input of the first neural network (the generator network of the GAN) is Gaussian white noise, and the output is the second channel sample information This second sample information and Gaussian white noise are input into the signal through channel module. Exemplarily, in the Figure 16a signal through channel module, the transmitted signal is multiplied by the second sample information , and the product result is added to the Gaussian white noise, and the result is used as the input of the receiving end. Through machine learning methods, the receiving end neural network (e.g., constellation demapping neural network) and the transmitting end neural network (constellation mapping neural network) can be trained end-to-end.

[0222] Exemplarily, when the first channel sample information is the second reference signal (y) introduced above, a schematic structural diagram of the first device training the second neural network and / or the third neural network according to the second channel sample information is shown in Figure 16b . The input of the first neural network (the generator network of the GAN) is Gaussian white noise and the transmitted signal x, and the output of the first neural network is the second reference signal Connect the first neural network between the receiving end and the transmitting end, and through machine learning methods, the receiving end neural network (e.g., constellation demapping neural network) and the transmitting end neural network (constellation mapping neural network) can be trained end-to-end.

[0223] In this embodiment, after the first device generates the second channel sample information according to the first neural network, the first device trains the second neural network and / or the third neural network according to the second channel sample information. The second neural network and / or the third neural network are used for the first device and the second device to transmit target information.

[0224] Optionally, the first device trains the second neural network and / or the third neural network according to the first channel sample information and the second channel sample information.

[0225] The second neural network and the third neural network are described separately below:

[0226] In the embodiments of the present application, the second neural network is applied to the transmitting end, and it can be understood that the second neural network performs the transmission processing of data, for example: as Figure 9The constellation mapping neural network shown. Optionally, the second neural network may perform operations such as rate matching and / or OFDM modulation. Among them, rate matching means that the bits to be transmitted are repeated and / or punctured to match the carrying capacity of the physical channel, so that the bit rate required by the transmission format is achieved during channel mapping. OFDM modulation mainly realizes the migration of the baseband spectrum to the radio frequency band, thereby realizing the wireless transmission function. Exemplarily, the second neural network may perform one or more functions among source coding (such as scrambling and / or signal compression, etc.), channel coding, constellation mapping, OFDM modulation, precoding, and / or filtering.

[0227] The network types of the second neural network include but are not limited to: fully connected neural network, radial basis function (RBF) neural network, CNN, recurrent neural network, Hopfield neural network, restricted Boltzmann machine, or deep belief network, etc. The second neural network may be any one of the above neural networks, or the second neural network may also be a combination of the above multiple neural networks, and no limitation is made here.

[0228] In the embodiments of this application, for the convenience of description, the first device is used as the sending end for illustration, that is, the second neural network is applied to the first device. It can be understood that the second device can also be used as the sending end, that is, the second neural network is applied to the second device.

[0229] In the embodiments of this application, the third neural network is applied to the receiving end, which can be understood as the third neural network receiving and processing the data from the sending end. For example: as Figure 9 The constellation demapping neural network shown. Specifically, the third neural network may perform source decoding (such as descrambling and / or signal decompression), channel decoding, demodulation of constellation mapping, OFDM demodulation, signal detection, and / or equalization, etc.

[0230] The network types of the third neural network include but are not limited to: fully connected neural network, radial basis neural network, convolutional neural network, recurrent neural network, Hopfield neural network, restricted Boltzmann machine, or deep belief network, etc. The third neural network may be any one of the above neural networks, or the third neural network may also be a combination of the above multiple neural networks, and no limitation is made here.

[0231] It should be noted that the second neural network and the third neural network may use the same neural network, or different neural networks, or partially the same neural networks, and no limitation is made here.

[0232] In the embodiments of the present application, for the convenience of description, the second device is taken as the receiving end for illustration, that is, the third neural network is applied to the second device. It can be understood that the first device can also be used as the receiving end, that is, the third neural network is applied to the first device.

[0233] The second neural network and / or the third neural network are used to transmit target information between the first device and the second device. Based on the above description, it can be understood that when the second neural network and the third neural network perform different processes, the target information is different.

[0234] Exemplarily, when the second neural network is a constellation mapping neural network and the third neural network is a constellation demapping neural network, the target information is a modulation symbol sequence.

[0235] Exemplarily, when the second neural network is a CSI compression neural network and the third neural network is a CSI decompression neural network, the target information is the compressed CSI or the CSI.

[0236] Exemplarily, when the second neural network is a channel coding neural network and the third neural network is a channel decoding neural network, the target information is the bit information to be transmitted before coding or the bit information to be transmitted after coding.

[0237] Exemplarily, when the second neural network is an OFDM modulation neural network and the third neural network is an OFDM demodulation neural network, the target information is an OFDM symbol.

[0238] Exemplarily, when the second neural network is a precoding neural network and the third neural network is an equalization neural network, the target information is a transmission signal, and the transmission signal reaches the receiving end after being transmitted through the channel.

[0239] The first device can train the second neural network alone, the first device can also train the third neural network alone, and the first device can also jointly train the second neural network and the third neural network.

[0240] Exemplarily, when the second neural network is a precoding neural network at the sending end (such as the first device), the first device can train the second neural network alone. The function of the precoding neural network is to generate the transmission weighting coefficients of each transmitting antenna according to the spatial characteristics of the channel, so that multiple transmission data streams can be spatially isolated, improving the received signal-to-interference-plus-noise ratio. The input of the precoding neural network is the CSI of the channel, and the output is the precoding weights on each transmitting antenna. That is, the target information of the second neural network is the precoding weights on each transmitting antenna.

[0241] Exemplarily, when the third neural network is the equalizer neural network of the receiving end (such as the second device), the first device can train the third neural network alone. The function of the equalizer neural network is to cancel the distortion effect caused by the channel on the propagated signal and restore the signal transmitted by the transmitting end. The input of the equalizer neural network is the CSI of the channel and the second reference signal of the receiving end, and the output is the restored signal after equalization. That is, the target information of the third neural network is the restored signal after equalization.

[0242] The first device can train the second neural network alone using the second channel sample information generated by the first neural network, or can train the third neural network alone using the second channel sample information and the locally generated transmission data, and send the information of the trained third neural network to the second device for use. The first device can also jointly train the second and third neural networks to achieve the optimal end-to-end communication performance.

[0243] When the first device finishes training the third neural network, it can execute step 1007. The completion of the training of the third neural network includes, but is not limited to: the number of training samples used in the training reaches a preset threshold. For example, if the number of second channel sample information for training the third neural network reaches 45,000, then the third neural network is regarded as trained. Or, the number of training times reaches a preset threshold. For example, if the number of training times of the third neural network reaches 30,000, then the third neural network is regarded as trained.

[0244] Optionally, 1007. The first device sends the information of the third neural network to the second device.

[0245] In this embodiment, when the first device finishes training the third neural network, the first device sends the information of the third neural network to the second device. The information of the third neural network includes, but is not limited to: the weights of the neural network, the activation function of the neurons, the number of neurons in each layer of the neural network, the inter-layer cascade relationship of the neural network, and / or the network type of each layer in the neural network. Exemplarily, the information of the third neural network can indicate the same or different activation functions for different neurons, without limitation.

[0246] Optionally, when a third neural network is pre-configured (or pre-defined) in the second device, the information of the third neural network may also be the model variation of the third neural network. The model variation includes, but is not limited to: the weights of the neural network that has changed, the activation function that has changed, the number of neurons in each layer of the neural network that has changed, the inter-layer cascading relationship of the neural network that has changed, and / or the network type of each layer in the neural network that has changed. Exemplarily, the pre-configuration may be that the access network device pre-configures for the terminal device through signaling, and the pre-definition may be pre-defined by the protocol. For example, the protocol pre-defines that the third neural network in the terminal device is neural network A.

[0247] The first device sends the information of the third neural network to the second device through signaling.

[0248] In the embodiments of the present application, the signaling includes, but is not limited to: broadcast messages (such as master information block (MIB)), system messages (such as system information block (SIB)), radio resource control (RRC) signaling, medium access control control element (MAC CE), and / or downlink control information (DCI). Among them, MAC CE and / or DCI may be common messages of multiple devices, or specific messages specific to a certain device (such as the second device).

[0249] The second device applies the third neural network in the second device according to the information of the third neural network.

[0250] After the first device finishes training the second neural network, the first device applies the second neural network.

[0251] When the second neural network is pre-configured (or pre-defined) in the first device, the first device updates the local second neural network according to the training result of step 1006. When the third neural network is pre-configured in the second device, the second device updates the local third neural network according to the information of the third neural network in step 1007. When there is no third neural network in the second device, the information of the third neural network in step 1007 may include the complete third neural network. The second device configures the trained third neural network locally according to the information of the third neural network.

[0252] After that, if the target is downlink information, the first device can use the second neural network to process the target information (i.e., the input of the second neural network) or obtain the target information (i.e., the output of the second neural network). The second device can use the third neural network to process the target information (i.e., the input of the third neural network) or recover the target information (i.e., the output of the third neural network). Exemplarily, the first device uses the second neural network to obtain modulation symbols, and the second device uses the third neural network to recover bits from the received modulation symbols.

[0253] If the target is uplink information, the second device can use the third neural network to process the target information (i.e., the input of the third neural network) or obtain the target information (i.e., the output of the third neural network). The first device can use the second neural network to process the target information (i.e., the input of the second neural network) or recover the target information (i.e., the output of the second neural network). Exemplarily, the second device uses the third neural network to compress CSI and obtains compressed CSI; the first device uses the second neural network to recover CSI from the received compressed CSI.

[0254] In the embodiments of this application, the second device (such as a terminal device) can send a relatively small amount of first channel sample information to the first device (such as an access network device). The first device trains a first neural network based on the first channel sample information. The first neural network is used to infer the second channel sample information. The first device trains the second neural network and / or the third neural network based on the second channel sample information, or based on the second channel sample information and the first channel sample information. The second neural network and / or the third neural network are used for the first device and the second device to transmit target information. Through this method, the signaling overhead in the air interface can be effectively reduced. At the same time, it can adapt to the channel environment. The trained second neural network and third neural network are closer to the actual channel environment, improving communication performance. The speed of training the second neural network and the third neural network is also greatly improved.

[0255] Similar to the above Figure 10The relevant steps for training the first neural network, the relevant training steps for the second neural network, and / or the relevant training steps for the third neural network can alternatively be implemented by other devices independent of the first device. In the embodiments of the present application, for the convenience of description, such other devices are referred to as the third device. The third device may be the AI node, mobile edge computing device, or cloud server described above, without limitation. The reference signal pool learned by the third device may be agreed upon in the protocol after offline learning, or sent to the first device through the interface between the third device and the first device, or forwarded to the first device through other network elements, without limitation. The samples required when the third device performs model training, such as the first channel sample information, may be directly or indirectly sent to the third device by the first device, or directly or indirectly sent to the third device by the second device, without limitation.

[0256] Please refer to Figure 11 , Figure 11 which is another flowchart of a method for training a neural network proposed in the embodiments of the present application. The method for training a neural network proposed in the embodiments of the present application includes:

[0257] Optionally, 1101. The first device sends a first reference signal to the second device.

[0258] 1102. The second device performs channel estimation based on the received first reference signal from the first device and determines the first channel sample information.

[0259] Optionally, 1103. The second device sends the first channel sample information to the first device.

[0260] Steps 1101-1103 are the same as the aforementioned steps 1001-1003 and will not be elaborated here.

[0261] 1104. The first device sends the first channel sample information to the third device.

[0262] In this embodiment, when the first device receives the first channel sample information from the second device, the first device sends the first channel sample information to the third device.

[0263] 1105. The third device trains the first neural network based on the first channel sample information.

[0264] 1106. The third device uses the first neural network to infer the second channel sample information.

[0265] Optionally, 1106 can be replaced with: The third device sends information of the first neural network to the first device. The first device uses the first neural network to infer and obtain second channel sample information. Optionally, the first device sends the second channel sample information to the third device.

[0266] 1107. The third device trains the second neural network and / or the third neural network according to the second channel sample information.

[0267] Steps 1105 - 1107 are similar to the foregoing steps 1004 - 1006 and will not be elaborated here.

[0268] 1108. The third device sends information of the second neural network and / or information of the third neural network to the first device.

[0269] In this embodiment, after the third device trains the third neural network, the third device can send the information of the third neural network to the second device through the first device. Optionally, the third device can also send the information of the third neural network to the first device through other devices (such as other access network devices). Optionally, if there is a direct communication link between the third device and the first device, the third device can also directly send the information of the third neural network to the first device.

[0270] After the third device trains the second neural network, the third device can send the information of the second neural network to the second device.

[0271] 1109. The first device sends the information of the third neural network to the second device.

[0272] Step 1109 is the same as the foregoing step 1007 and will not be elaborated here.

[0273] Optionally, Figure 11 The shown embodiment can be replaced with: The entity for training the first neural network is different from the entity for training the second neural network and / or the third neural network. For example, the former is the first device and the latter is the third device; or, the former is the third device and the latter is the first device.

[0274] In the embodiments of the present application, by training the neural network in the above manner, the signaling overhead of the air interface is effectively reduced, and at the same time, it can adapt to the channel environment. The trained second neural network and third neural network are closer to the actual channel environment, improving the communication performance. The speed of training the second neural network and the third neural network is also greatly improved. The steps related to neural network training can be executed by other devices (the third device), effectively reducing the computing load and power consumption of the first device (such as an access network device).

[0275] Similar to the above Figure 10 or Figure 11Optionally, the steps related to training the first neural network can alternatively be implemented by a second device (e.g., a terminal device).

[0276] Please refer to Figure 12 , Figure 12 FIG. is another schematic flowchart of a method for training a neural network proposed in an embodiment of the present application. In this method, the first neural network is trained by a second device. The method for training a neural network proposed in an embodiment of the present application includes:

[0277] Optionally, 1201. The second device sends capability information to the first device.

[0278] In this embodiment, similar to the foregoing Figures 10 - 11 shown embodiment, taking the first device as an access network device and the second device as a terminal device as an example for illustration. The first device may be an access network device, a chip in the access network device, a module or circuit in the access network device, etc., and the second device may be a terminal device, a chip in the terminal device, a module or circuit in the terminal, etc. The first device may also be a terminal device, a chip in the terminal device, a module or circuit in the terminal, etc., and the second device may also be an access network device or a chip in the access network device, a module or circuit in the access network device, etc., which is not limited herein.

[0279] In an embodiment of the present application, for ease of description, the first neural network that has not been trained using the first channel sample information, or the first neural network before update, may be referred to as a reference neural network. After training the reference neural network using the first channel sample information, the first neural network is obtained.

[0280] This reference neural network is similar to the reference neural network in step 1004 above. The second device starts training this reference neural network using the first channel sample information, thereby training the first neural network.

[0281] In step 1201, the second device sends capability information to the first device, and this capability information is used to indicate one or more of the following information:

[0282] 1) Whether it supports using a neural network to replace or implement the functions of a communication module;

[0283] The communication module includes, but is not limited to: an OFDM modulation module, an OFDM demodulation module, a constellation mapping module, a constellation demapping module, a channel coding module, a channel decoding module, a precoding module, an equalization module, an interleaving module, and / or a deinterleaving module.

[0284] 2) Whether it supports the network type of the first neural network, or the network types supported by the first neural network;

[0285] For example, the capability information indicates whether the second device supports VAE or whether it supports GAN. As another example, the capability information indicates that the second device supports GAN, supports VAE, supports both GAN and VAE, or supports neither GAN nor VAE.

[0286] 3) Whether it supports the network type of the third neural network, or indicates the supported network type of the third neural network;

[0287] Among them, the network type of the third neural network includes one or more of the following: fully connected neural network, radial basis neural network, convolutional neural network, recurrent neural network, Hopfield neural network, restricted Boltzmann machine, or deep belief network, etc. The third neural network can be any one of the above neural networks, or the third neural network can also be a combination of the above multiple neural networks, and there is no restriction here.

[0288] 4) Whether it supports receiving information of a reference neural network through signaling, where the reference neural network is a first neural network not trained with the first channel sample information or is used to train the first neural network;

[0289] 5) Whether it supports receiving information of the third neural network through signaling;

[0290] 6) The pre-configured reference neural network stored;

[0291] For example: If there is already a predefined or pre-configured reference neural network in the second device, the identifier or index of the reference neural network can be carried in the capability information.

[0292] Optionally, the identifier or index of other predefined or pre-configured neural networks already existing in the second device can also be carried in the capability information.

[0293] 7) The memory space available for storing the first neural network and / or the third neural network;

[0294] 8) The computing power information available for running the neural network;

[0295] The computing power information refers to the computing ability information for running the neural network, such as including the operation speed of the processor and / or the amount of data that the processor can process, etc.

[0296] 9) The location information of the second device.

[0297] For example, report the longitude, latitude, and / or altitude of the second device. As another example, report the channel environment of the second device, such as: office, subway station, park, or street.

[0298] The capability information includes, but is not limited to, the above information. For example, the capability information may further include: the type of channel sample information supported by the second device for training the first neural network. Exemplarily, the channel sample information supported by the second device (terminal device) for training the first neural network is CSI, or the channel sample information supported by the second device for training the first neural network is the second reference signal (y).

[0299] Step 1201 is an optional step. When step 1201 is executed, the second device may send the capability information when initially accessing the first device, or the second device may send the capability information to the first device periodically, without limitation here.

[0300] Optionally, after the first device receives the capability information from the second device, when the capability information indicates that the reference neural network is not configured (not stored) in the second device. Then the first device may send the information of the reference neural network to the second device. The information of the reference neural network includes: the weights of the neural network, the activation function of the neurons, the number of neurons in each layer of the neural network, the inter-layer cascade relationship of the neural network, and / or the network type of each layer in the neural network. The information of the reference neural network may be referred to as the specific information of the reference neural network. Among them, the activation functions of different neurons may be the same or different, without limitation.

[0301] Optionally, 1202. The first device sends a first reference signal to the second device.

[0302] Same as step 1001 above, which will not be elaborated here.

[0303] Optionally, 1203. The second device performs channel estimation based on the received first reference signal from the first device to determine the first channel sample information.

[0304] Same as step 1002, which will not be elaborated here.

[0305] 1204. The second device trains the first neural network based on the first channel sample information.

[0306] The method by which the second device trains the first neural network based on the first channel sample information is similar to the method by which the first device trains the first neural network based on the first channel sample information in 1004. It will not be elaborated here.

[0307] Exemplarily, the second device trains the first neural network based on the first channel sample information and the reference neural network.

[0308] Optionally, the second device may determine the reference neural network according to any one of the following methods 1 to 3:

[0309] Method 1: The information of the reference neural network is agreed upon by the protocol, or the second device determines the information of the reference neural network.

[0310] In a possible implementation, the protocol agrees upon (pre - defines) the information of a reference neural network, that is, the specific information of the reference neural network is agreed upon.

[0311] In the embodiments of the present application, the information of the reference neural network includes: the weights of the neural network, the neuron activation function, the number of neurons in each layer of the neural network, the inter - layer cascade relationship of the neural network, and / or the network type of each layer in the neural network. The information of this reference neural network can also be referred to as the specific information of the reference neural network. Among them, the activation functions of different neurons can be the same or different, without limitation.

[0312] In a possible implementation, the protocol agrees upon the information of multiple reference neural networks. Each reference neural network corresponds to a channel environment. Among them, the channel environment includes one or more of the following: office, indoor shopping mall, subway station, urban street, square, and suburb, etc. As described above, the second device reports the location information of the second device to the first device, then the first device can determine the channel environment where the second device is located, so as to obtain the reference neural network corresponding to this channel environment. For example, the access network device can maintain a map information, and the channel environment corresponding to different location information marks is marked in this map information. For example: This map information is shown in Table 1:

[0313] Table 1

[0314] Location information / {longitude, latitude, altitude} Channel environment {x1, y1, z1} Office {x2, y2, z2} Subway station {x3, y3, z3} Park {x4, y4, z4} Street

[0315] Optionally, if the second device reports the location information of the second device, that is, the channel environment of the second device, to the first device through the capability information, then the first device can determine the information of the reference neural network corresponding to this channel environment.

[0316] In a possible implementation, the second device determines the information of the reference neural network, that is, determines the specific information of the reference neural network. As described above, the second device can report the information of this reference neural network to the first device through the capability information.

[0317] Method 2: The reference neural network is indicated by the first device for the second device.

[0318] In a possible implementation, the first device sends the information of the reference neural network to the second device, that is, the first device sends the specific information of the reference neural network to the second device.

[0319] In a possible implementation, the protocol stipulates information of multiple candidate reference neural networks, and each candidate reference neural network corresponds to a unique index. The first device sends an identifier to the second device via signaling, and this identifier is used to indicate the index of the reference neural network configured for the second device among the multiple candidate reference neural networks.

[0320] In a possible implementation, the first device configures information of multiple candidate reference neural networks for the second device via a first signaling, that is, configures the specific information of the multiple candidate reference neural networks. Among them, each candidate reference neural network corresponds to a unique index. The first device sends an identifier to the second device via a second signaling, and this identifier is used to indicate the index of the reference neural network configured for the second device among the multiple candidate reference neural networks.

[0321] In a possible implementation, as described above, the second device can send information of multiple candidate reference neural networks stored or possessed by the second device to the first device via capability information. Among them, each candidate reference neural network corresponds to a unique index. The first device sends an identifier to the second device via signaling, and this identifier is used to indicate the index of the reference neural network configured for the second device among the multiple candidate reference neural networks.

[0322] In a possible implementation, the first device sends the type of the reference neural network to the second device via signaling. The protocol stipulates information of various types of reference neural networks, or the second device independently determines information of various types of reference neural networks.

[0323] 1205. The second device sends information of the first neural network to the first device.

[0324] After the second device trains the first neural network, the second device sends information of the first neural network to the first device.

[0325] Optionally, the information of the first neural network can be the specific information of the first neural network. This specific information can include: the weights of the neural network, the activation function of the neurons, the number of neurons in each layer of the neural network, the inter-layer cascade relationship of the neural network, and / or the network type of each layer in the neural network. Among them, the activation functions of different neurons can be the same or different, without limitation.

[0326] Optionally, the information of the first neural network can also be the model change amount of the first neural network relative to the reference neural network. The model change amount includes but is not limited to: the weights of the neural network that have changed, the activation function that has changed, the number of one or more layers of neurons in the neural network that have changed, the inter-layer cascade relationship of the neural network that has changed, and / or the network type of one or more layers in the neural network that have changed. When the information of the first neural network is the model change amount, the overhead of air interface signaling can be effectively reduced.

[0327] 1206. The first device determines the first neural network according to the information of the first neural network.

[0328] The first device determines the first neural network according to the information of the first neural network from the second device.

[0329] Optionally, when the information of the first neural network is the specific information of the first neural network, the first device determines the first neural network according to the specific information.

[0330] Optionally, when the information of the first neural network is the model change amount, the first device determines the first neural network according to the model change amount and the reference neural network.

[0331] 1207. The first device infers the second channel sample information according to the first neural network.

[0332] 1208. The first device trains the second neural network and / or the third neural network according to the second channel sample information.

[0333] 1209. The first device sends the information of the third neural network to the second device.

[0334] Steps 1207 - 1209 are the same as the aforementioned steps 1005 - 1007 and will not be elaborated here.

[0335] In this embodiment, the second device trains the first neural network according to the first channel sample information and the reference neural network.

[0336] In this embodiment, the second device (such as a terminal device) trains the first neural network according to the first channel sample information. The first neural network is used to infer the second channel sample information. The first device trains the second neural network and / or the third neural network according to the second channel sample information. The second neural network and / or the third neural network are used for the first device and the second device to transmit target information. It can effectively reduce the air interface signaling overhead, and at the same time can adapt to the channel environment. The trained second neural network and third neural network are closer to the actual channel environment, improving the communication performance. The speed of training the second neural network and the third neural network is also greatly improved. Assigning the computing task of training the first neural network to the second device can effectively reduce the computing burden of the first device (such as an access network device) and improve the communication performance.

[0337] In the embodiments provided by the present application above, the methods provided by the embodiments of the present application are introduced from the perspectives of the first device, the second device, the third device, and the interactions between them. To implement each function in the methods provided by the embodiments of the present application above, the first device, the second device, and the third device may include a hardware structure and / or software module, and implement the above functions in the form of a hardware structure, a software module, or a combination of a hardware structure and a software module. Whether a certain function among the above functions is executed in the form of a hardware structure, a software module, or a combination of a hardware structure and a software module depends on the specific application and design constraints of the technical solution.

[0338] Please refer to Figure 17 , Figure 17 FIG. is a schematic diagram of a communication device provided by an embodiment of the present application. The communication device 1700 includes a transceiver module 1710 and a processing module 1720, where:

[0339] Optionally, the communication device 1700 is used to implement the functions of the first device in the above method embodiment.

[0340] In a possible implementation, the transceiver module 1710 is configured to send a first reference signal to the second device;

[0341] The transceiver module 1710 is further configured to receive first channel sample information from the second device;

[0342] The processing module 1720 is configured to determine a first neural network. The first neural network is trained according to the first channel sample information. The first neural network is used to infer second channel sample information. The second channel sample information is used to train a second neural network and / or a third neural network. The second neural network and / or the third neural network are used for the first device to transmit target information to the second device.

[0343] In a possible implementation, the transceiver module 1710 is configured to send a first reference signal to the second device;

[0344] The transceiver module 1710 is further configured to receive information of the first neural network from the second device;

[0345] The processing module 1720 is configured to determine a second neural network and / or a third neural network. The second neural network and / or the third neural network are used for the first device to transmit target information to the second device. The second neural network and / or the third neural network are trained according to the second channel sample information. The second channel sample information is inferred by the first neural network.

[0346] Optionally, the communication device 1700 is used to implement the functions of the second device in the above method embodiment.

[0347] In a possible implementation, the processing module 1720 is configured to perform channel estimation based on a first reference signal received from a first device, and determine first channel sample information;

[0348] The transceiver module 1710 is configured to send the first channel sample information to the first device;

[0349] The transceiver module 1710 is further configured to receive information of a third neural network from the first device, where the third neural network is used for the second device and the first device to transmit target information.

[0350] In a possible implementation, the processing module 1720 is configured to perform channel estimation based on a first reference signal received from a first device, and determine first channel sample information;

[0351] The processing module 1720 is further configured to determine a first neural network, where the first neural network is trained according to the first channel sample information;

[0352] The transceiver module 1710 is configured to send the information of the first neural network to the first device.

[0353] Optionally, the communication device 1700 is used to implement the functions of the third device in the above method embodiments.

[0354] In a possible implementation, the transceiver module 1710 is configured to receive first channel sample information from the first device;

[0355] The processing module 1720 is configured to determine a first neural network, where the first neural network is trained according to the first channel sample information, and the first neural network is used to infer second channel sample information, and the second channel sample information is used to train a second neural network and / or a third neural network, and the second neural network and / or the third neural network are used for the first device and the second device to transmit target information.

[0356] Optionally, the above communication device may further include a storage unit, where the storage unit is configured to store data and / or instructions (which may also be referred to as code or programs). Each of the above units may interact with or be coupled to the storage unit to implement corresponding methods or functions. For example, the processing unit 1720 may read data or instructions from the storage unit, so that the communication device implements the methods in the above embodiments. The coupling in the embodiments of the present application is an indirect coupling or communication connection between devices, units or modules, and may be electrical, mechanical or other forms for information interaction between devices, units or modules.

[0357] In the embodiments of the present application, the division of modules is illustrative, and is only a logical function division. In actual implementation, there may be other division methods. In addition, in each embodiment of the present application, each functional module may be integrated in a processor, may exist alone physically, or two or more modules may be integrated in one module. The above integrated modules may be implemented in the form of hardware or in the form of software functional modules.

[0358] In one example, the units in any of the above communication devices may be one or more integrated circuits configured to implement the above methods. For example: one or more application specific integrated circuits (ASICs), or, one or more digital signal processors (DSPs), or, one or more field programmable gate arrays (FPGAs), or a combination of at least two of these integrated circuit forms. Again, when the units in the communication device can be implemented in the form of a processing element scheduler, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processors that can call programs. Again, these units may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0359] The technical solutions provided by the embodiments of this application can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, an AI node, an access network device, a terminal device, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that includes one or more integrated available media. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a digital video disc (DVD)), or a semiconductor medium, etc.

[0360] In the embodiments of this application, on the premise of no logical contradiction, the embodiments can refer to each other. For example, the methods and / or terms between method embodiments can refer to each other, for example, the functions and / or terms between device embodiments can refer to each other, and for example, the functions and / or terms between device embodiments and method embodiments can refer to each other.

[0361] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalent technologies, this application is also intended to include these modifications and variations.

Claims

1. A training method for a neural network, characterized in that, applied to a first device, the method includes: sending a first reference signal to a second device; receiving first channel sample information from the second device; determining a first neural network, the first neural network being trained according to the first channel sample information, the first neural network being used to infer second channel sample information, the second channel sample information being used to train a second neural network and / or a third neural network, the second neural network and / or the third neural network being used for the first device and the second device to transmit target information.

2. The method according to claim 1, characterized in that, the first channel sample information is also used to train the second neural network and / or the third neural network.

3. The method according to any one of claims 1-2, characterized in that, the method further includes: the first device sending information of the third neural network to the second device.

4. The method according to any one of claims 1-3, characterized in that, the first channel sample information includes channel state information CSI and / or a second reference signal, the second reference signal being the first reference signal propagated through the channel.

5. The method according to any one of claims 1-4, characterized in that, the first neural network is a generative adversarial network or a variational autoencoder.

6. The method according to any one of claims 1-5, characterized in that, the first reference signal includes a demodulation reference signal DMRS or a channel state information reference signal CSI-RS.

7. The method according to any one of claims 1-6, characterized in that, the sequence type of the first reference signal includes a ZC sequence or a gold sequence.

8. A training method for a neural network, characterized in that, applied to a second device, the method includes: performing channel estimation according to a first reference signal received from a first device to determine first channel sample information; sending the first channel sample information to the first device; receiving information of a third neural network from the first device, the third neural network being used for the second device and the first device to transmit target information.

9. The method according to claim 8, characterized in that, the first channel sample information includes channel state information CSI and / or a second reference signal, the second reference signal being the first reference signal received by the second device.

10. A communication device, characterized in that, configured to implement the method according to any one of claims 1-9.

Images