Communication method, communication device and storage medium

Abstract

The embodiment of the invention discloses a communication method, a communication device and a storage medium, which are applied to the technical field of communication and are used for improving model performance. The method comprises the steps that M pieces of output data are obtained, the M pieces of output data are obtained by processing first input data through N models, any one of the N models is used for reasoning the first input data, M is a positive integer, and N is a positive integer smaller than or equal to M; k pieces of output data are determined based on the M pieces of output data, wherein K is a positive integer. According to the embodiment of the invention, the K output data can be determined based on the M output data, so that the pre-training model can keep stable output or performance, and the robustness of the pre-training model is improved.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to a communication method, communication device and storage medium. Background Technology

[0002] In recent years, large models have been widely used in various fields of computer science, including natural language processing, computer vision, speech recognition, and recommender systems. Currently, large models have significantly improved performance in multimodal tasks, and one possible way to enhance model performance is to utilize scaling laws to simultaneously increase both the training dataset and the number of model parameters. This has led to an explosive growth in the number of parameters in large models.

[0003] Currently, pre-trained models can be deployed on the network side. Since the pre-trained model is a large model, large model fine-tuning techniques can be used to decompose the weight parameters of the large model into low-rank values, thereby enabling personalized inference.

[0004] However, in personalized reasoning, how to improve the robustness of model training or model inference is an urgent problem to be solved. Summary of the Invention

[0005] This application provides a communication method, communication device, and storage medium for improving model performance.

[0006] The first aspect of this application provides a communication method. Optionally, the executing entity of the method can be a first communication device, which can be a network device, a component or device applied to the network device (e.g., a processor, chip, or chip system), or a logic module or software capable of implementing all or part of the functions of the network device (e.g., a central unit (CU), a distributed unit (DU), or a radio unit (RU)). The first communication device can also be a terminal device, a component or device applied to the terminal device (e.g., a processor, chip, or chip system), or a logic module or software capable of implementing all or part of the functions of the terminal device. In this method, the first communication device acquires M output data, which are obtained by processing the first input data through N models. Any one of the N models is used to infer the first input data, where M is a positive integer and N is a positive integer less than or equal to M. The first communication device determines K output data based on the M output data, where K is a positive integer.

[0007] In this embodiment, the M output data acquired by the first communication device are obtained by processing the first input data through N models, and these N models are suitable for inference on the first input data. Specifically, these N models are pre-trained models used for inference. Based on the M output data, K output data are determined, thereby enabling the pre-trained models to maintain the stability of their output or performance, thus improving the robustness of the pre-trained models.

[0008] Based on the first aspect of this application, in some possible implementations, if at least one of the M output data does not satisfy the first condition, K output data are determined.

[0009] In this embodiment, when at least one of the M output data does not satisfy the first condition, the first communication device determines that an abnormal state has occurred during the processing of the N models, thereby determining the K output data. In other words, the first communication device can improve the robustness of the N models by monitoring whether an abnormal state has occurred.

[0010] Based on the first aspect of this application, in some possible implementations, at least one output data does not satisfy the first condition, including:

[0011] At least one output data point has a first parameter lower than a first preset value, where the first parameter indicates the degree of correlation between the at least one output data point and the first input data; and / or, at least one output data point has a second parameter that satisfies a first preset condition, where the second parameter indicates the diversity of the at least one output data point. The degree of correlation between the at least one output data point and the first input data can be understood as the inference quality of the N models.

[0012] In this embodiment, by limiting the first condition, when the inference quality of the N models is low or the diversity of the at least one output data is insufficient, the first communication device determines that an abnormal state has occurred in the processing of the N models.

[0013] Based on the first aspect of this application, in some possible implementations, the K output data satisfy any one of the following:

[0014] K output data are contained within M output data; among the M output data, K output data satisfy the first condition;

[0015] The K output data are obtained by reprocessing the first input data through N models;

[0016] The K output data are obtained by processing the first input data through P models. The P models are different from the N models, where P is an integer greater than or equal to N.

[0017] The K output data are obtained by processing the second input data through N models, and the second input data is different from the first input data; or,

[0018] The K output data are obtained by processing the second input data through P models.

[0019] In this embodiment, the first communication device can determine K output data by filtering output data that does not meet the first condition, reprocessing based on the first input data, switching the pre-trained model, or switching the input data, thereby obtaining output data with high inference quality or more diversity, and thus improving the robustness of the N models.

[0020] Based on the first aspect of this application, in some possible implementations, a first communication device receives first information, which is used to determine P models.

[0021] In this embodiment, the first communication device can determine P models based on the first information, so that the receiver of the first information can obtain the model information of the first model through the first information, so that the sender or receiver of the first information can manage N models based on the first information.

[0022] Based on the first aspect of this application, in some possible implementations, the first communication device receives second information, the second information being configuration information for requesting second input data; the first communication device then sends the configuration information.

[0023] In this embodiment, the first communication device can receive second information, enabling the sender of the second information to obtain the configuration information of the second input data through the second information, so that the sender of the second information can obtain the second input data based on the second information.

[0024] Based on the first aspect of this application, in some possible implementations, the first communication device sends third information, which is used to indicate K output data.

[0025] In this embodiment, the first communication device can send third information, enabling the recipient of the third information to obtain the K output data through the third information, so that the recipient of the third information can process based on the K output data.

[0026] Based on the first aspect of this application, in some possible implementations, the first communication device acquires X output data, which are obtained by processing K output data through Y models, or the X output data are label data generated from third input data, and any one of the Y models is used to score the third input data, where X is a positive integer and Y is a positive integer less than or equal to X; the first communication device determines Z output data based on the X output data, where Z is a positive integer.

[0027] In this embodiment, the first communication device can process K output data using Y models to obtain Z output data. The second communication device acquires X output data, which are obtained by processing the K output data through the Y models, or the X output data are label data generated from the K output data. When the X output data are obtained by processing the K output data through the Y models, the Y models are used to determine the personalized target output; that is, the Y models can be personalized models. When the X output data are label data generated from the K output data, the label data is used to train the personalized model. Therefore, determining Z output data based on the X output data improves the robustness of the personalized model's training or inference.

[0028] Based on the first aspect of this application, in some possible implementations, the first communication device receives fourth information, which is used to determine Q models, and any one of the Q models is used to score K data.

[0029] In this embodiment, the first communication device can receive fourth information, enabling the sender of the fourth information to determine Q models, so that the sender of the fourth information can score the K outputs based on the Q models.

[0030] Based on the first aspect of this application, in some possible implementations, the Z output data satisfy any one of the following:

[0031] Z output data are contained within X output data; among the X output data, Z output data satisfy the second condition;

[0032] Z output data are obtained by reprocessing K output data through Y models;

[0033] Z output data are obtained by processing K output data through Q models. At least one of the Q models is different from Y models, where Q is an integer greater than or equal to Y.

[0034] Z output data are obtained by processing G output data through Y models. At least one of the G output data is different from K output data. G output data are obtained by processing the third input data through N models. G is an integer greater than or equal to Y.

[0035] Z output data are obtained by processing G output data through Q models; or,

[0036] Z output data are obtained by processing G output data. G output data are the label data corresponding to K output data. At least one of the G output data is different from the K output data. G is an integer greater than or equal to Y.

[0037] In this embodiment, the first communication device can determine Z output data by filtering output data that does not meet the second condition, reprocessing based on the third input data, switching the pre-trained model, or switching the input data, thereby obtaining output data with high inference quality or more diversity, and thus improving the robustness of the Y models.

[0038] A second aspect of this application provides a communication method. Optionally, the execution entity of this method can be a second communication device, which can be a network device, a component or device applied to a network device (e.g., a processor, chip, or chip system), or a logic module or software (e.g., CU, DU, or RU) capable of implementing all or part of the functions of the network device. The second communication device can also be a terminal device, a component or device applied to a terminal device (e.g., a processor, chip, or chip system), or a logic module or software capable of implementing all or part of the functions of the terminal device. In this method, the second communication device acquires X output data, which are obtained by processing a fourth input data through Y models, or the X output data are label data generated from the fourth input data. Any one of the Y models is used to score the fourth input data, where X is a positive integer and Y is a positive integer less than or equal to X. The second communication device determines Z output data based on the X output data, where Z is a positive integer.

[0039] In this embodiment, the X output data acquired by the second communication device are obtained by processing the fourth input data through Y models, or the X output data are label data generated from the fourth input data. When the X output data are obtained by processing the fourth input data through Y models, the Y models are used to determine the personalized target output, that is, the Y models can be personalized models. When the X output data are label data generated from the fourth input data, the label data is used to train the personalized model. Determining Z output data based on the X output data improves the robustness of the personalized model's training or inference.

[0040] Based on the second aspect of this application, in some possible implementations, the second communication device determines Z output data when at least one of the X output data does not satisfy the second condition.

[0041] In this embodiment, when at least one of the X output data does not satisfy the second condition, the second communication device determines that an abnormal state has occurred during the processing of the Y models, and thus determines the Z output data. In other words, the second communication device can improve the robustness of the Y models by monitoring whether an abnormal state has occurred.

[0042] Based on the second aspect of this application, in some possible implementations, at least one output data does not satisfy the second condition, including:

[0043] At least one of the X output data satisfies the second preset condition in its third parameter, which indicates at least one of the distribution function, extreme values, mean, variance, or interval of the output data.

[0044] In this embodiment, by limiting the second condition, when the third parameter of at least one of the X output data does not meet the preset condition, the first communication device determines that an abnormal state has occurred in the processing of the Y models.

[0045] Z output data are contained within X output data; among the X output data, Z output data satisfy the second condition;

[0046] The Z output data are obtained by reprocessing the fourth input data through Y models;

[0047] Z output data are obtained by processing the fourth input data through Q models. The Q models are different from the Y models, where Q is an integer greater than or equal to Y.

[0048] Z output data are obtained by processing the fifth input data through Y models. The fifth input data is different from the fourth input data.

[0049] The Z output data are obtained by processing the fifth input data through Q models; or,

[0050] The Z output data are obtained by processing the fifth input data, which is the label data corresponding to the fourth input data. At least one of the output data in the fifth input data is different from the fourth input data, and G is an integer greater than or equal to Y.

[0051] In this embodiment, the second communication device can determine Z output data by filtering output data that does not meet the second condition, reprocessing based on the fourth input data, switching the pre-trained model, or switching the input data, thereby obtaining output data with high inference quality or more diversity, and thus improving the robustness of the Y models.

[0052] Based on the second aspect of this application, in some possible implementations, the second communication device determines Z output data when at least one of the X output data does not satisfy the second condition.

[0053] In this embodiment, when at least one of the X output data does not satisfy the second condition, the second communication device determines that an abnormal state has occurred during the processing of generating the Y models or labels, and thus determines the Z output data. In other words, the first communication device can improve the robustness of the Y models by monitoring whether abnormal states occur during the training or inference process of the Y models.

[0054] Based on the second aspect of this application, in some possible implementations, at least one output data does not satisfy the second condition, including:

[0055] At least one of the Z output data satisfies the second preset condition in its third parameter, which indicates at least one of the distribution function, extreme values, mean, variance, or interval of the output data.

[0056] In this embodiment, by limiting the second condition, when the overall score of the reasoning results of the Y models is low or cannot be distinguished as good or bad, or when the true value distribution of the X label data deviates from the benchmark value, the second communication device determines that an abnormal state has occurred in the processing of the generation of the Y models or labels.

[0057] Based on the second aspect of this application, in some possible implementations, the second communication device receives third information, which is used to indicate K output data, the K output data being fourth input data.

[0058] In this embodiment, the sender of the third information can use a pre-trained model to obtain K output data and instruct the second communication device through the third information, so that the second communication device can determine the fourth input data according to the third information, thereby determining Z output data in X output data, and thus improving the robustness of the Y models.

[0059] Based on the second aspect of this application, in some possible implementations, the second communication device sends first information, which is used to determine P models, and the P models are used to infer fourth input data, where P is a positive integer.

[0060] In this embodiment, the second communication device can send first information so that the receiver of the first information determines P models based on the first information, and the receiver of the first information can deduce fourth input data based on the P models, so that the second communication device can obtain X output data based on the fourth input data.

[0061] Based on the second aspect of this application, in some possible implementations, the second communication device sends second information, which is used to request configuration information for the second input data. The second input data is used to process N models to obtain fourth input data, and any one of the N models is used to perform reasoning on the second input data, where N is a positive integer; the second communication device receives the configuration information.

[0062] In this embodiment, the second communication device can send second information, enabling the recipient of the second information to determine the configuration information of the second input data based on the second information, so that the second communication device can obtain the second input data based on the second information.

[0063] Based on the second aspect of this application, in some possible implementations, the second communication device sends fourth information, which is used to determine Q models.

[0064] Based on the second aspect of this application, in some possible implementations, the second communication device sends a third message, which is used to determine P models.

[0065] In this embodiment, the second communication device can receive a third message, enabling the second communication device to instruct the recipient of the third message to determine P models, thereby enabling the recipient of the third message to obtain K output data based on the P models, and thus enabling the second communication device to acquire the K output data as the fourth input data.

[0066] Based on the second aspect of this application, in some possible implementations, the second communication device sends a fourth message, the fourth message being used to request configuration information, the configuration information being configuration information of the second input data; the second communication device receives the configuration information.

[0067] In this embodiment, the second communication device can receive a third message, enabling the second communication device to request configuration information of the second input data from the recipient of the third message through the fourth message. Thus, the recipient of the third message can determine the configuration information of the second input data based on the fourth message, thereby enabling the second communication device to collect the second input data based on the configuration information of the second input data.

[0068] Based on the first or second aspect of this application, in some possible embodiments, a first communication device sends third information indicating K output data; a second communication device determines X output data based on the K output data, the X output data being tag data of the K output data. The first communication device receives the X output data.

[0069] In this embodiment, the first communication device can send third information, enabling the second communication device to obtain the K output data through the third information, so that the second communication device can process the K output data to obtain X output data, thereby enabling the first communication device to obtain the X output data.

[0070] A third aspect of this application provides a communication device, comprising:

[0071] The processing module is used to obtain M output data. The M output data are obtained by processing the first input data through N models. Any one of the N models is used to perform inference on the first input data. M is a positive integer and N is a positive integer less than or equal to M.

[0072] The processing module is also used to determine K output data based on M output data, where K is a positive integer.

[0073] In one possible implementation, the processing module is used to acquire M output data, including: the processing module receives the M output data through the interface module.

[0074] In another possible implementation, the processing module is further configured to determine K output data based on M output data, including: the processing module is specifically configured to determine K output data when at least one of the M output data does not satisfy a first condition.

[0075] In another possible implementation, the interface module is also used to receive first information, which is used to determine P models.

[0076] In another possible implementation, the interface module is also used to receive second information, which is configuration information for requesting the second input data;

[0077] The interface module is also used to send configuration information.

[0078] In another possible implementation, the interface module is also used to send a third message, which indicates K output data.

[0079] In another possible implementation, the interface module is also used to receive fourth information, which is used to determine Q models, any one of which is used to score K data points.

[0080] In another possible implementation, the processing module is also used to determine Z output data based on X output data, where Z is a positive integer; wherein, the X output data are obtained by processing K output data through Y models, or, the X output data are label data generated from K output data, and any one of the Y models is used to score the K output data, where X is a positive integer and Y is a positive integer less than or equal to X.

[0081] In another possible implementation, the processing module is further configured to determine Z output data based on X output data, including: the processing module is specifically configured to determine Z output data when at least one of the X output data does not satisfy the second condition.

[0082] A fourth aspect of this application provides a communication device, comprising:

[0083] The processing module is used to obtain X output data, which are obtained by processing the third input data through Y models, or the X output data are label data generated from the third input data. Any one of the Y models is used to score the third input data, where X is a positive integer and Y is a positive integer less than or equal to X.

[0084] The processing module is also used to determine Z output data based on X output data, where Z is a positive integer.

[0085] In one possible implementation, the processing module is used to acquire X output data, including: the processing module receives the X output data through the interface module.

[0086] In another possible implementation, the processing module is further configured to determine Z output data based on X output data, including: the processing module is specifically configured to determine Z output data when at least one of the X output data does not satisfy the second condition.

[0087] In another possible implementation, the interface module is also used to receive third information, which indicates third input data.

[0088] In another possible implementation, the interface module is also used to send first information, which is used to determine P models, and the P models are used to reason about the first input data to obtain the third input data, where P is a positive integer.

[0089] In another possible implementation, the interface module is also used to send second information, which is used to request configuration information for the second input data. The second input data is used to process the second input data through N models to obtain the third input data. Any one of the N models is used to perform inference on the second input data, where N is a positive integer.

[0090] Receive configuration information.

[0091] A fifth aspect of this application provides a communication device, which may be a first communication device or a second communication device, or a component (e.g., a processor, chip, or chip system) applied to the first or second communication device, or a logic module or software (e.g., CU, DU, or RU) capable of implementing all or part of the functions of the first or second communication device. The communication device includes:

[0092] A processor for executing a program that causes the communication device to perform the method as described in the first or second aspect of the foregoing and any possible implementation thereof.

[0093] Optionally, the communication device further includes a memory, and the processor is coupled to the memory; the memory is used to store programs.

[0094] The sixth aspect of this application provides a chip or chip system including at least one processor and a communication interface, the communication interface and at least one processor being interconnected via a line, the at least one processor being used to run computer programs or instructions to perform the communication method described in any of the possible implementations of the first or second aspect.

[0095] The communication interface in the chip can be an input / output interface, pins, or circuits.

[0096] In one possible implementation, the chip or chip system described above in this application further includes at least one memory storing instructions. The memory can be an internal storage unit of the chip, such as a register or cache, or it can be a storage unit of the chip itself, such as a read-only memory or random access memory.

[0097] The seventh aspect of this application provides a communication system, including a communication device that performs the first aspect and any possible implementation thereof, and a communication device that performs the second aspect and any possible implementation thereof.

[0098] An eighth aspect of this application provides a computer-readable storage medium including instructions that, when executed on a computer, cause the computer to perform the method described in the first aspect above, or cause the computer to perform the method described in the second aspect above.

[0099] The ninth aspect of this application provides a computer program product containing instructions that, when run on a computer, cause the computer to perform the method described in the first aspect above, or cause the computer to perform the method described in the second aspect above. Attached Figure Description

[0100] Figures 1a to 1cA schematic diagram of the communication system provided in this application;

[0101] Figures 2a to 2g This is a schematic diagram of the AI ​​processing involved in this application;

[0102] Figure 3 A schematic diagram of an embodiment of the communication method provided in this application;

[0103] Figures 4 to 7 Some possible implementations of the communication method provided in this application;

[0104] Figure 8 A schematic diagram of another embodiment of the communication method provided in this application;

[0105] Figures 9 to 15 Some possible implementations of the communication method provided in this application;

[0106] Figures 16 to 20 A schematic diagram of the communication device provided in this application. Detailed Implementation

[0107] First, some terms used in the embodiments of this application will be explained to facilitate understanding by those skilled in the art.

[0108] (1) Terminal device: can be a wireless terminal device that can receive network device scheduling and instruction information. The wireless terminal device can be a device that provides voice and / or data connectivity to the user, or a handheld device with wireless connection function, or other processing device connected to a wireless modem.

[0109] Terminal devices can communicate with one or more core networks or the Internet via a radio access network (RAN). Terminal devices can be mobile terminal devices, such as mobile phones (or "cellular" phones), computers, and data cards. For example, they can be portable, pocket-sized, handheld, computer-embedded, or vehicle-mounted mobile devices that exchange voice and / or data with the RAN. Examples include personal communication service (PCS) phones, cordless phones, session initiation protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs), tablets, and computers with wireless transceiver capabilities. Wireless terminal equipment can also be called subscriber unit, subscriber station, mobile station (MS), remote station, access point (AP), remote terminal, access terminal, user terminal, user agent, subscriber station (SS), customer premises equipment (CPE), terminal, user equipment (UE), mobile terminal (MT), etc.

[0110] By way of example and not limitation, in this embodiment, the terminal device can also be a wearable device. Wearable devices, also known as wearable smart devices or smart wearable devices, are a general term for devices that utilize wearable technology to intelligently design and develop everyday wearables, such as glasses, gloves, watches, clothing, and shoes. Wearable devices are portable devices that are worn directly on the body or integrated into the user's clothing or accessories. Wearable devices are not merely hardware devices, but also achieve powerful functions through software support, data interaction, and cloud interaction. Broadly speaking, wearable smart devices include those that are feature-rich, large in size, and can achieve complete or partial functions without relying on a smartphone, such as smartwatches or smart glasses, as well as those that focus on a specific type of application function and require the use of other devices such as smartphones, such as various smart bracelets, smart helmets, and smart jewelry for vital sign monitoring.

[0111] Terminals can also be drones, robots, devices in device-to-device (D2D) communication, vehicles to everything (V2X) communication, virtual reality (VR) terminal devices, augmented reality (AR) terminal devices, wireless terminals in industrial control, wireless terminals in self-driving, wireless terminals in telemedicine or telehealth services, wireless terminals in smart grids, wireless terminals in transportation safety, wireless terminals in smart cities, wireless terminals in smart homes, etc.

[0112] Furthermore, terminal devices can also be terminal devices in future communication systems beyond the fifth generation (5G) (such as 5G Advanced communication systems) or in future evolved public land mobile networks (PLMNs). For example, 5G Advanced networks can further expand the form and function of 5G communication terminals, including but not limited to vehicles, cellular network terminals (integrating satellite terminal functions), drones, and Internet of Things (IoT) devices.

[0113] In this embodiment, the terminal device can also obtain artificial intelligence (AI) services provided by the network device. Optionally, the terminal device can also have AI processing capabilities.

[0114] (2) Network equipment: This can be equipment within a wireless network. For example, network equipment can be a RAN node (or device) that connects terminal devices to the wireless network, and can also be called a base station. Currently, some examples of RAN equipment include: base station, evolved NodeB (eNodeB), gNB (gNodeB) in 5G communication systems, transmission reception point (TRP), evolved Node B (eNB), radio network controller (RNC), Node B (NB), home base station (e.g., home-evolved Node B, or home Node B, HNB), base band unit (BBU), or wireless fidelity (Wi-Fi) access point (AP), etc. In addition, in a network architecture, network equipment can include central unit (CU) nodes, distributed unit (DU) nodes, or RAN equipment including both CU and DU nodes.

[0115] Optionally, the RAN node can also be a macro base station, micro base station, indoor station, relay node, donor node, or a radio controller in a cloud radio access network (CRAN) scenario. The RAN node can also be a server, wearable device, vehicle, or in-vehicle equipment. For example, the access network equipment in vehicle-to-everything (V2X) technology can be a roadside unit (RSU).

[0116] In another possible scenario, multiple RAN nodes collaborate to assist the terminal in achieving wireless access, with different RAN nodes each implementing a portion of the base station's functions. For example, RAN nodes can be CUs, DUs, CUs (control plane, CP), CUs (user plane, UP), or radio units (RUs). CUs and DUs can be configured separately or included in the same network element, such as a baseband unit (BBU). RUs can be included in radio equipment or radio units, such as remote radio units (RRUs), active antenna units (AAUs), radio heads (RHs), or remote radio heads (RRHs).

[0117] In different systems, CU (or CU-CP and CU-UP), DU, or RU may have different names, but those skilled in the art will understand their meaning. For example, in an open access network (open RAN, O-RAN, or ORAN) system, CU can also be called O-CU (open CU), DU can also be called O-DU, CU-CP can also be called O-CU-CP, CU-UP can also be called O-CU-UP, and RU can also be called O-RU. For ease of description, this application uses CU, CU-CP, CU-UP, DU, and RU as examples. Any of the units among CU (or CU-CP, CU-UP), DU, and RU in this application can be implemented through software modules, hardware modules, or a combination of software modules and hardware modules.

[0118] Communication between access network devices and terminal devices follows a specific protocol layer structure. This protocol layer may include a control plane protocol layer and a user plane protocol layer. The control plane protocol layer may include at least one of the following: radio resource control (RRC) layer, packet data convergence protocol (PDCP) layer, radio link control (RLC) layer, media access control (MAC) layer, or physical (PHY) layer, etc. The user plane protocol layer may include at least one of the following: service data adaptation protocol (SDAP) layer, PDCP layer, RLC layer, MAC layer, or physical layer, etc.

[0119] The correspondence between network elements and their achievable protocol layer functions in the ORAN system can be found in Table 1 below.

[0120] Table 1

[0121] ORAN network elements 3GPP protocol layer functions O-CU-CP RRC+PDCP-Control Plane (PDCP-C) O-CU-UP SDAP+PDCP - User Plane (PDCP-U) O-DU RLC+MAC+PHY-high O-RU PHY-low

[0122] Network devices can be other devices that provide wireless communication functions for terminal devices. The embodiments of this application do not limit the specific technology or form of the network device. For ease of description, the embodiments of this application are not limited.

[0123] Network equipment may also include core network equipment, such as the Mobility Management Entity (MME), Home Subscriber Server (HSS), Serving Gateway (S-GW), Policy and Charging Rules Function (PCRF), and Public Data Network Gateway (PDN gateway or P-GW) in 4th generation (4G) networks; and access and mobility management function (AMF), user plane function (UPF), or session management function (SMF) in 5G networks. Furthermore, this core network equipment may also include other core network equipment in 5G networks and next-generation networks of 5G networks.

[0124] In this embodiment of the application, the network device may also have network nodes with AI capabilities, which can provide AI services to terminals or other network devices. For example, it may be an AI node, computing node, RAN node with AI capabilities, or core network element with AI capabilities on the network side (access network or core network).

[0125] In this application embodiment, the device for implementing the function of the network device can be the network device itself, or it can be a device capable of supporting the network device in implementing the function, such as a chip system. This device can be disposed within the network device. In the technical solutions provided in this application embodiment, the example of a network device being used to implement the function of the network device is used to describe the technical solutions provided in this application embodiment.

[0126] (3) Configuration and Pre-configuration: In this application, both configuration and pre-configuration are used. Configuration refers to the network device / server sending configuration information or parameter values ​​to the terminal via messages or signaling, so that the terminal can determine communication parameters or transmission resources based on these values ​​or information. Pre-configuration is similar to configuration; it can be parameter information or parameter values ​​negotiated in advance between the network device / server and the terminal device, or it can be parameter information or parameter values ​​used by the base station / network device or terminal device as specified in standard protocols, or it can be parameter information or parameter values ​​pre-stored in the base station / server or terminal device. This application does not limit this.

[0127] Furthermore, these values ​​and parameters can be changed or updated.

[0128] (4) The terms "system" and "network" in the embodiments of this application can be used interchangeably. "Multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, or B exists alone, where A and B can be singular or plural. The character " / " generally indicates that the related objects before and after are in an "or" relationship. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, "at least one of A, B and C" includes A, B, C, AB, AC, BC or ABC. And, unless otherwise specified, the ordinal numbers such as "first" and "second" mentioned in the embodiments of this application are used to distinguish multiple objects and are not used to limit the order, sequence, priority or importance of multiple objects.

[0129] (5) In the embodiments of this application, "send" and "receive" indicate the direction of signal transmission. For example, "send information to XX" can be understood as the destination of the information being XX, which may include sending directly through the air interface or sending indirectly through the air interface by other units or modules. "Receive information from YY" can be understood as the source of the information being YY, which may include receiving directly from YY through the air interface or receiving indirectly from YY through the air interface by other units or modules. "Send" can also be understood as the "output" of the chip interface, and "receive" can also be understood as the "input" of the chip interface.

[0130] In other words, sending and receiving can occur between devices, such as between network devices and terminal devices, or within a device, such as between components, modules, chips, software modules, or hardware modules within the device via buses, wiring, or interfaces.

[0131] It is understandable that information may undergo necessary processing, such as encoding and modulation, between the source and destination, but the destination can understand the valid information from the source. Similar statements in this application can be interpreted in a similar way and will not be elaborated further.

[0132] (6) In the embodiments of this application, "instruction" may include direct instruction and indirect instruction, as well as explicit instruction and implicit instruction. The information indicated by a certain piece of information (as described below, the instruction information) is called the information to be instructed. In the specific implementation process, there are many ways to indicate the information to be instructed, such as, but not limited to, directly indicating the information to be instructed, such as the information to be instructed itself or its index. It can also indirectly indicate the information to be instructed by indicating other information, where there is an association between the other information and the information to be instructed; or it can only indicate a part of the information to be instructed, while the other parts of the information to be instructed are known or pre-agreed upon. For example, the instruction can be implemented by using a pre-agreed (e.g., protocol predefined) arrangement order of various information, thereby reducing the instruction overhead to a certain extent. This application does not limit the specific method of instruction. It is understood that for the sender of the instruction information, the instruction information can be used to indicate the information to be instructed, and for the receiver of the instruction information, the instruction information can be used to determine the information to be instructed.

[0133] In this application, unless otherwise specified, the same or similar parts between the various embodiments can be referred to each other. In the various embodiments of this application, and in the various methods / designs / implementations within each embodiment, unless otherwise specified or logically conflicting, the terminology and / or descriptions between different embodiments and between the various methods / designs / implementations within each embodiment are consistent and can be mutually referenced. The technical features in different embodiments and the various methods / designs / implementations within each embodiment can be combined to form new embodiments, methods, or implementations based on their inherent logical relationships. The following descriptions of the embodiments of this application do not constitute a limitation on the scope of protection of this application.

[0134] This application can be applied to long-term evolution (LTE) systems, new radio (NR) systems, or future communication systems beyond 5G. These communication systems include at least one network device and / or at least one terminal device.

[0135] Please see Figure 1a This is a schematic diagram of the architecture of the communication system 1000 used in an embodiment of this application. Figure 1aAs shown, the communication system may include a radio access network (RAN) 100. Optionally, the communication system 1000 may also include a core network 200 and an Internet 300. The RAN 100 includes at least one RAN node (e.g., ...). Figure 1a 110a and 110b, collectively referred to as 110, may also include at least one terminal (such as...). Figure 1a RAN100, denoted as RAN100, comprises RAN nodes 120a-120j, collectively referred to as RAN120. RAN100 may also include other RAN nodes, such as wireless relay equipment and / or wireless backhaul equipment. Figure 1a (Not shown in the image). Terminal 120 connects wirelessly to RAN node 110, and RAN node 110 connects wirelessly or via a wired connection to core network 200. The core network equipment in core network 200 and RAN node 110 in RAN 100 can be independent physical devices, or they can be the same physical device integrating the logical functions of core network equipment and RAN nodes. Terminals can connect to each other, and RAN nodes can connect to each other, via wired or wireless connections.

[0136] by Figure 1a Taking the communication system shown as an example, in addition to performing communication-related services, different devices (including network devices and terminal devices, and / or terminal devices and terminal devices) may also perform AI-related services.

[0137] like Figure 1b As shown, taking a network device as a base station as an example, a base station can perform communication-related services and AI-related services with one or more terminal devices, and different terminal devices can also perform communication-related services and AI-related services.

[0138] like Figure 1c As shown, taking terminal devices including TVs and mobile phones as an example, TVs and mobile phones can also perform communication-related services and AI-related services.

[0139] The technical solution provided in this application can be applied to wireless communication systems (e.g.) Figure 1a , Figure 1b or Figure 1cThe system shown, for example, the communication system provided in this application, can incorporate AI network elements to implement some or all AI-related operations. AI network elements can also be called AI nodes, AI devices, AI entities, AI modules, AI models, or AI units, etc. The AI ​​network element can be built into a network element within the communication system. For example, an AI network element can be an AI module built into: access network equipment, core network equipment, cloud server, or operation, administration, and maintenance (OAM) to implement AI-related functions. The OAM can act as the network management system for the core network equipment and / or the access network equipment. Alternatively, the AI ​​network element can also be a network element independently set up in the communication system. Optionally, the terminal or its built-in chip can also include an AI entity to implement AI-related functions.

[0140] Optionally, in communication systems, AI application cases may include, but are not limited to: channel state information (CSI) feedback enhancement, beam management enhancement, positioning accuracy enhancement, network energy saving, load balancing, and mobility optimization. These will be explained below.

[0141] 1. Enhanced CSI feedback

[0142] Channel quality information (CSI) is the channel attribute of a communication link, reported by the terminal device to the network device. By reporting this information, the terminal device can select an appropriate modulation and coding scheme (MCS) to adapt to changing wireless channels. For example, the terminal device might perform channel estimation based on the received channel state information-reference signal (CSI-RS) and then feed back the CSI-RS to the network device. This information serves as input to the network device's model, enabling AI model training. Applying AI to CSI feedback enhancement can reduce overhead, improve accuracy, and enhance predictive capabilities.

[0143] CSI-RS feedback enhancement may include at least one sub-function, such as: CSI compression, CSI prediction, and CSI-RS configuration signaling reduction. CSI compression may further include CSI compression in at least one domain: spatial, time, and frequency.

[0144] 2. Enhanced Beam Management

[0145] Enhanced beam management primarily aims to discover the strongest transmit / receive beam pairs. AI-based sparse beam prediction can improve accuracy. This can be achieved through both network-side and terminal-side AI sparse beam prediction, based on AI training and inference. Taking terminal-side AI sparse beam prediction as an example, the pre-trained AI model on the terminal device can be provided by the network or pre-stored on the terminal device. During training, the network device scans all possible beams and then reports the transmit beam pattern to the terminal device. Once training is complete, the network device only needs to scan a small subset of beams, and the terminal device then feeds back the inference results. AI-based beam management can achieve beam prediction in, for example, the temporal and / or spatial domains, reducing overhead and latency and improving beam selection accuracy.

[0146] Beam management enhancements may include at least one sub-function, such as: beam scan matrix prediction and optimal beam prediction.

[0147] 3. Enhanced positioning accuracy

[0148] In line-of-sight (LOS) or non-line-of-sight (NLOS) scenarios, AI-based positioning can improve positioning accuracy with a smaller number of TRP antennas. Positioning enhancement can include at least one sub-function, such as: positioning enhancement based on access network devices, positioning enhancement based on positioning management function network elements, and positioning enhancement based on terminal devices.

[0149] 4. Network energy saving

[0150] Network energy conservation can be achieved through cell activation / deactivation, load reduction, coverage improvement, or other RAN setting adjustments. AI technology can be used to optimize energy-saving decisions by leveraging data collected within the RAN network. AI algorithms can predict energy efficiency and load status for the next cycle, which can be used to assist in cell activation / deactivation decisions to save energy. Based on the predicted load, the system can dynamically configure energy-saving strategies to maintain a balance between system performance and energy efficiency, and reduce energy consumption.

[0151] 5. Load balancing

[0152] Load balancing can distribute the load evenly between cells and across different areas within a cell, or transfer some traffic from congested cells, or offload users across a single cell, carrier, or access standard, thereby improving network performance. Using AI models to enhance load balancing performance—such as inputting various measurements and feedback from terminal devices and network nodes, as well as historical data—can provide a higher quality user experience and increase system capacity.

[0153] 6. Mobility Management

[0154] Mobility management is a solution that ensures service continuity for mobile devices by minimizing dropped calls, radio link failures (RLFs), unnecessary handovers, and ping-pong effects. AI can enhance mobility management by, for example, reducing the probability of unexpected events, predicting device location / mobility / performance, and routing traffic.

[0155] It should be understood that the definitions of the above technical terms are merely illustrative. For example, as technology continues to develop, the scope of the above definitions may also change, and the embodiments of this application are not intended to limit the scope.

[0156] For example, an AI function may include multiple AI sub-functions.

[0157] Alternatively, AI application cases are also referred to as AI application scenarios or AI functions.

[0158] As described above regarding AI application examples, AI can be widely used to improve network performance in areas such as CSI feedback enhancement, beam management, positioning accuracy enhancement, energy saving, mobility enhancement, and load balancing. AI models can typically be deployed on the network side and / or the terminal device side. The training of AI models relies on the collection of training data, which can come from measurements and feedback from the terminal devices.

[0159] The following is a brief introduction to the concepts that may be involved in this application.

[0160] AI can endow machines with human-like intelligence, for example, allowing them to use computer hardware and software to simulate certain intelligent human behaviors. To achieve artificial intelligence, machine learning methods can be employed. In machine learning, machines learn (or train) a model using training data. This model represents the mapping between inputs and outputs. The learned model can be used for reasoning (or prediction), that is, it can be used to predict the output corresponding to a given input. This output can also be called the reasoning result (or prediction result).

[0161] Machine learning can include supervised learning, unsupervised learning, and reinforcement learning. Unsupervised learning can also be called learning without supervision.

[0162] Supervised learning, based on collected sample values ​​and labels, uses machine learning algorithms to learn the mapping relationship between sample values ​​and labels, and then expresses this learned mapping relationship using an AI model. The process of training the machine learning model is the process of learning this mapping relationship. During training, sample values ​​are input into the model to obtain the model's predicted values, and the model parameters are optimized by calculating the error between the model's predicted values ​​and the sample labels (ideal values). After the mapping relationship is learned, it can be used to predict new sample labels. The mapping relationship learned in supervised learning can include linear or non-linear mappings. Based on the type of label, the learning task can be divided into classification tasks and regression tasks.

[0163] Unsupervised learning relies on collected sample values ​​to discover inherent patterns within the samples themselves. One type of unsupervised learning algorithm uses the samples themselves as supervisory signals, meaning the model learns the mapping relationship from sample to sample; this is called self-supervised learning. During training, model parameters are optimized by calculating the error between the model's predictions and the samples themselves. Self-supervised learning can be used for signal compression and decompression recovery applications; common algorithms include autoencoders and generative adversarial networks.

[0164] Reinforcement learning, unlike supervised learning, is a type of algorithm that learns problem-solving strategies through interaction with the environment. Unlike supervised and unsupervised learning, reinforcement learning problems do not have explicit "correct" action labels. The algorithm needs to interact with the environment to obtain reward signals from the environment, and then adjust its decision actions to obtain a larger reward signal value. For example, in downlink power control, the reinforcement learning model adjusts the downlink transmission power of each user based on the total system throughput feedback from the wireless network, aiming to achieve a higher system throughput. The goal of reinforcement learning is also to learn the mapping relationship between the environment state and a better (e.g., optimal) decision 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." Reinforcement learning training is achieved through iterative interaction with the environment.

[0165] Neural networks (NNs) are a specific model in machine learning techniques. According to the general approximation theorem, neural networks can theoretically approximate any continuous function, thus enabling them to learn arbitrary mappings. Traditional communication systems rely on extensive expert knowledge to design communication modules, while deep learning communication systems based on neural networks can automatically discover hidden pattern structures from large datasets, establish mapping relationships between data, and achieve performance superior to traditional modeling methods.

[0166] The idea behind neural networks comes from the neuronal structure of the brain. For example, each neuron performs a weighted summation of its input values ​​and outputs the result through an activation function.

[0167] like Figure 2a The diagram shown is a schematic representation of a neuron structure. Assume the neuron's input is x = [x0, x1, ..., x...]. n The weights corresponding to each input are w = [w0, w1, ..., w] n ], where n is a positive integer, w i and x i It can be any possible type, such as a decimal, an integer (e.g., 0, a positive integer, or a negative integer), or a complex number. i As x i The weights are used to assign weights to x. i Weighting is applied. The bias for the weighted sum of the input values ​​is, for example, b. Activation functions can take many forms. Suppose the activation function of a neuron is: y = f(z) = max(0, z), then the output of that neuron is: For example, if the activation function of a neuron is y = f(z) = z, then the output of that neuron is: Here, b can be any possible type, such as a decimal, an integer (e.g., 0, a positive integer, or a negative integer), or a complex number. The activation functions of different neurons in a neural network can be the same or different.

[0168] Furthermore, neural networks generally consist of multiple layers, each of which may include one or more neurons. Increasing the depth and / or width of a neural network can improve its expressive power, providing more powerful information extraction and abstract modeling capabilities for complex systems. The depth of a neural network can refer to the number of layers it includes, and the number of neurons in each layer can be called the width of that layer. In one implementation, a neural network includes an input layer and an output layer. The input layer processes the received input information through neurons and passes the processing result to the output layer, which then obtains the output of the neural network. In another implementation, a neural network includes an input layer, hidden layers, and an output layer. The input layer processes the received input information through neurons and passes the processing result to the hidden layer. The hidden layer calculates the received processing result and passes the calculation result to the output layer or the next adjacent hidden layer, ultimately obtaining the output of the neural network. A neural network may include one hidden layer or multiple sequentially connected hidden layers, without limitation.

[0169] Neural networks, for example, are deep neural networks (DNNs). Depending on how the network is constructed, DNNs can include feedforward neural networks (FNNs), convolutional neural networks (CNNs), and recurrent neural networks (RNNs).

[0170] Figure 2b This is a schematic diagram of a Free-Nearest Neural Network (FNN). A key characteristic of FNNs is that neurons in adjacent layers are completely connected pairwise. This characteristic makes FNNs typically require a large amount of storage space, leading to high computational complexity.

[0171] CNNs are neural networks specifically designed to process data with a grid-like structure. For example, time-series data (e.g., discrete sampling along a time axis) and image data (e.g., two-dimensional discrete sampling) can both be considered grid-like data. CNNs do not use all the input information at once for computation; instead, they use a fixed-size window to extract a portion of the information for convolution operations, which significantly reduces the computational cost of model parameters. Furthermore, depending on the type of information extracted by the window (e.g., people and objects in an image represent different types of information), each window can use different convolution kernels, allowing CNNs to better extract features from the input data.

[0172] Recurrent Neural Networks (RNNs) are a type of neural network that utilizes feedback time-series information. The input to an RNN includes the current input value and its own output value from the previous time step. RNNs are suitable for acquiring temporally correlated sequence features, and are applicable to applications such as speech recognition and channel coding / decoding.

[0173] In the model training process described above, a loss function can be defined. The loss function describes the difference between the model's output value and the ideal target value. The loss function can be expressed in various forms, and there are no restrictions on its specific form. The model training process can be viewed as follows: by adjusting some or all of the model's parameters, the value of the loss function is made to be less than a threshold or to meet the target requirement.

[0174] A model can also be called an AI model, a rule, or other names. An AI model can be considered a specific method for implementing AI functions. An AI model represents the mapping relationship or function between the model's input and output. AI functions can include one or more of the following: data collection, model training (or model learning), model information dissemination, model inference (or model reasoning, inference, or prediction, etc.), model monitoring or model validation, or inference result publication, etc. AI functions can also be called AI (related) operations or AI-related functions.

[0175] The implementation process of the neural network will be described below with reference to the accompanying drawings.

[0176] 1. Fully connected neural network, also known as multilayer perceptron (MLP).

[0177] like Figure 2c As shown, an MLP consists of an input layer (left side), an output layer (right side), and multiple hidden layers (middle). Each layer of an MLP contains several nodes, called neurons. Neurons in adjacent layers are connected pairwise.

[0178] Optionally, considering neurons in two adjacent layers, the output h of a neuron in the next layer is the weighted sum of all neurons x in the previous layer connected to it, processed by an activation function, and can be expressed as:

[0179] h = f(wx + b).

[0180] Where w is the weight matrix, b is the bias vector, and f is the activation function.

[0181] Alternatively, the output of the neural network can be recursively expressed as:

[0182] y = f z (w z f z-1 (…)+b z ).

[0183] Where z is the index of the neural network layer, z is greater than or equal to 1 and z is less than or equal to Z, where Z is the total number of layers in the neural network.

[0184] In other words, a neural network can be understood as a mapping from an input data set to an output data set. Neural networks are typically initialized randomly; the process of obtaining this mapping from random values ​​w and b using existing data is called training the neural network.

[0185] Optionally, the training process may involve evaluating the output of the neural network using a loss function.

[0186] like Figure 2d As shown, the error can be backpropagated, and the neural network parameters (including w and b) can be iteratively optimized using gradient descent until the output of the loss function reaches its minimum value. Figure 2d The term "relative advantage (e.g., optimal advantage)" is used. This is understandable. Figure 2d The neural network parameters corresponding to the "better points (e.g., the best points)" in the data can be used as neural network parameters in the trained AI model information.

[0187] Alternatively, the gradient descent process can be represented as:

[0188]

[0189] Where θ represents the parameters to be optimized (including w and b), L is the loss function, and η is the learning rate, controlling the step size of gradient descent. This represents the differentiation operation. This indicates taking the derivative of θ with respect to L.

[0190] Alternatively, the backpropagation process may utilize the chain rule for partial derivatives.

[0191] like Figure 2e As shown, the gradient of the parameters in the previous layer can be recursively calculated from the gradient of the parameters in the next layer, and can be expressed as:

[0192]

[0193] Among them, w ij Let s be the weight of the connection between node j and node i. i The weighted sum of the inputs at node i.

[0194] 2. Federated Learning (FL).

[0195] The concept of federated learning effectively addresses the current challenges in the development of artificial intelligence. While fully protecting user data privacy and security, it enables various edge devices and central servers to collaborate efficiently to complete the model's learning task.

[0196] like Figure 2f As shown, the FL architecture is currently the most widely used training architecture in the FL field, and the FedAvg algorithm is the foundational algorithm of FL. The FedAvg algorithm flow is roughly as follows:

[0197] (1) Initialize the model to be trained at the center end. And broadcast it to all clients.

[0198] (2) In the t∈[1,T] round, the client k∈[1,K] is based on the local dataset. For the received global model Perform E epochs of training to obtain the local training results. Report it to the central node. Figure 2f In the example shown, the local training results sent by distributed nodes n, k, and m are denoted as G, respectively. n G k G m .

[0199] (3) The central node collects local training results from all (or some) clients. Assume the set of clients uploading local models in round t is... The central server will use the number of samples from the corresponding client as weights to calculate the new global model. The specific update rule is as follows: Then the central end will send the latest version of the global model. The broadcast is sent to all clients for a new round of training.

[0200] (4) Repeat steps (2) and (3) until the model finally converges or the number of training rounds reaches the upper limit.

[0201] Optionally, in addition to reporting the local model, the client can also... It can also train local gradients The central node averages the local gradients reported by all clients and updates the global model based on this average gradient.

[0202] As can be seen in the FL framework, the dataset resides on distributed nodes (such as clients). These distributed nodes collect their local datasets, perform local training, and report the local results (model or gradients) to the central node. The central node itself may not have a dataset; it can be responsible for fusing the training results from the distributed nodes to obtain a global model, which is then distributed back to the distributed nodes.

[0203] 3. Decentralized learning.

[0204] like Figure 2g The diagram shows a fully distributed system without a central node. The design goal of a decentralized learning system is generally the goal f of each node. i The mean of (x), i.e. Where n is the number of distributed nodes, and x is the parameter to be optimized; in machine learning, x is the parameter of the machine learning model (such as a neural network). Each node utilizes local data and its local target f. i (x) Calculate the local gradient Then it is sent to its communicatively reachable neighboring nodes. Upon receiving the gradient information from its neighbor, any node can update the parameters x of its local model according to the following formula:

[0205]

[0206] in, This represents the parameters of the local model after the (k+1)th update (k is a natural number) in the i-th node. This represents the parameters of the local model for the i-th node after the k-th update (if k is 0, then it represents...). (where α is the parameter of the local model of the i-th node that is not involved in the update) k N represents the tuning coefficient. i It is the set of neighboring nodes of node i, |N i | represents the number of elements in the set of neighboring nodes of node i, that is, the number of neighboring nodes of node i. Through information interaction between nodes, the decentralized learning system will eventually learn a unified model.

[0207] The technical solution provided in this application can be applied to communication systems (e.g.) Figure 1a or Figure 1b or Figure 1c In a communication system (as shown in the diagram), communication nodes typically possess both signal transmission and reception capabilities and computational capabilities. Taking a network device with computational capabilities as an example, the network device's computational capabilities primarily provide computing power support for signal transmission and reception capabilities (e.g., processing signals for transmission and reception) to enable the network device to perform communication tasks with other communication nodes.

[0208] With the development of communication technology, communication systems can now perform not only traditional communication services but also new types of services, such as artificial intelligence (AI) services. Generally, systems capable of handling AI services, such as communication systems, can also be called AI systems. However, improving user experience within AI systems remains a pressing technical challenge.

[0209] In one possible implementation, using the scaling law to increase the complexity of the AI ​​model deployed on the communication device (e.g., increasing the number of parameters or the number of neural network layers in the model) can effectively improve the model's performance and thus enhance the user experience.

[0210] As an example of scaling law, an AI model deployed on a communication device can learn the knowledge of a highly complex pre-trained large model deployed in the cloud. The former can be called the student model and the latter can be called the teacher model. This can improve the performance of the student model (e.g., improve the inference accuracy and precision of the student model) through knowledge distillation (or knowledge transfer) to enhance the user experience.

[0211] As another example of scaling law, AI models deployed on communication devices can be trained with large amounts of personalized data, which can improve the performance of the AI ​​model (e.g., improve the inference accuracy and precision of the student model) to enhance the user experience.

[0212] However, in AI systems, when the output data of AI models deployed on communication devices exhibits abnormal conditions, such as poor quality or insufficient diversity of output data, there is a lack of means to detect and repair such abnormal conditions, resulting in poor robustness of model training.

[0213] To address the aforementioned problems, this application provides a communication method, communication device, and storage medium that processes the model's output data to obtain new output data, thereby improving the model's robustness. A detailed description will follow with reference to the accompanying drawings.

[0214] Please see Figure 3 This is a schematic diagram of an implementation of the communication method provided in this application, which includes the following steps.

[0215] It should be noted that in the following text Figure 3 In the following implementation examples, the method is illustrated using a first communication device and other communication devices (such as a second communication device) as the execution subjects of the interaction illustration. However, this application does not limit the execution subjects of the interaction illustration. For example, the communication device can be a communication equipment, or a chip, baseband chip, modem chip, system-on-chip (SoC) chip containing a modem core, system-in-package (SIP) chip, communication module, chip system, processor, logic module, or software in the communication equipment. Optionally, the communication equipment can be a terminal device or network device (e.g., access network equipment, access network element, core network element, or core network equipment).

[0216] As an example, the first communication device can be a network device and the second communication device can be a terminal device, or both the first and second communication devices can be network devices. For example, the network device can be an access network device or a communication device in an ORAN system (e.g., at least one of CU, DU, RU).

[0217] As another example, both the first and second communication devices are terminal devices, i.e. Figure 3 The proposed solution can be applied to sidelink communication scenarios.

[0218] 301. The first communication device acquires M output data, wherein the M output data are obtained by processing the first input data through N models, where N is a positive integer and M is an integer greater than or equal to N;

[0219] The first communication device can obtain M output data in various ways, which will be introduced below with some examples.

[0220] Implementation Example 1: The first communication device receives the M output data. Thus, the first communication device can obtain the M output data by receiving the M output data, that is, the N models used to process the first input data can be deployed on other communication devices, so that the first communication device does not need to deploy the N models (e.g., N pre-trained large models), which can save the storage space of the first communication device and reduce the processing complexity of the first communication device.

[0221] In one possible implementation, the first communication device sends first input data. Thus, the first communication device can also send the first input data to one or more communication devices that have deployed N models, enabling the one or more communication devices to process the first input data and obtain and send M output data.

[0222] Optionally, the first input data can be pre-configured data, so that the first communication device does not need to send the first input data, thereby reducing overhead.

[0223] In another possible implementation, the first communication device sends a request message requesting M output data. Thus, the first communication device can also, by sending the request message, enable the recipient of the request message (e.g., a second communication device) to provide M output data to the first communication device based on the request message.

[0224] In another possible implementation, the first communication device receives configuration information for the first input data, which corresponds to the request information. After receiving the request information, the recipient of the request information (e.g., the second communication device) can determine the configuration information for the first input data based on the request information and send the configuration information, enabling the first communication device to determine the first input data based on the configuration information.

[0225] Example 2: The first communication device processes the first input data based on the N models to obtain the M output data.

[0226] In one possible implementation, the first communication device receives first input data. Thus, by receiving the first input data, the first communication device can obtain M output data through processing the first input data.

[0227] Optionally, the first input data can be pre-configured data, or the first data can be data collected by the first communication device itself, in order to reduce transmission overhead.

[0228] It should be understood that the M output data are obtained by processing the first input data through N models. This can be understood as each of the N models processing the first input data and then obtaining one or more output data. The M output data may include one or more output data corresponding to each of the N models.

[0229] It should be understood that any one of the M output data points can be the model output or intermediate result of one of the N models. In other words, one or more output data points corresponding to each of the N models can include the model output and / or intermediate results of each model.

[0230] For example, a model may include one or more neural networks, and the model output of the model can be understood as part or all of the output of the last neural network in the one or more neural networks. For example, the model output may include part or all of the output of the last neural network. In addition, the intermediate result of the model can be understood as part or all of the output of at least one other neural network in the one or more neural networks except for the last neural network. For example, the intermediate result may include part or all of the output of each of the at least one neural network.

[0231] As an example, in the above scheme, any one of the N models is used to process the first input data to obtain one or more output data. That is, the model processing performed by any one model can include one or more of reasoning, prediction, derivation, recognition, and decision-making. Accordingly, any one model can be called a reasoning model, a prediction model, a pre-trained model for the above model processing, or a pre-trained large model for the above model processing, etc.

[0232] 302. The first communication device determines K output data based on M output data, where K is a positive integer; the M output data are different from the K output data, for example, the M output data are partially the same as the K output data; or, at least one of the K output data is different from any one of the M output data; or, any one of the K output data is different from any one of the M output data.

[0233] In one possible implementation, the first communication device may determine the K output data when at least one of the M output data does not satisfy a first condition.

[0234] Optionally, if at least one output data point does not satisfy the first condition, it can be understood as at least one of the M output data points having a first parameter lower than a first preset value. The first parameter indicates the correlation between at least one of the M output data points and the first input data. Alternatively, the first parameter can be used to indicate the inference quality of at least one of the M output data points. For example, the first parameter could be the mutual information of the input and output data, the accuracy of the N models, etc., and is not specifically limited here.

[0235] Optionally, the fact that at least one output data does not satisfy the first condition can be understood as at least one of the M output data satisfying the first preset condition in terms of its second parameter. The second parameter is used to indicate the diversity of at least one output data. For example, the second parameter satisfying the first preset condition may be: the distance between the output data and other data in a preset space is lower than a first threshold, or the similarity between the output data and other data in the preset space is higher than a second threshold. The specific details are not limited here.

[0236] As an example, the first communication device can judge at least one of the M output data and determine K output data.

[0237] As an example, the at least one output data can be determined by a monitor deployed on the first communication device. Specifically, the monitor can be a hardware device or a software device deployed on the first communication device, and no specific limitation is made here. In this embodiment, the monitor used to monitor the M output data is referred to as the M0 monitor.

[0238] Optionally, when the M0 monitor determines that at least one of the M output data does not meet the first condition, the M0 monitor sends a fifth message to the first communication device. The fifth message is used to indicate that at least one of the M output data does not meet the first condition, or the fifth message is used to indicate that at least one of the M output data does not meet the first condition.

[0239] In this embodiment of the application, by determining K output data that satisfy the first condition based on M output data during model training or model inference, the robustness of model training or model inference is improved.

[0240] The first communication device can obtain K output data in various ways, which will be explained below.

[0241] Method 1: The first communication device filters out the output data that does not meet the first condition from the M output data, obtaining K output data. That is, the first communication device filters out MK output data. In other words, the K output data are contained in the M output data; among the M output data, K output data satisfy the first condition.

[0242] Optionally, the first communication device may generate a data index when filtering the MK output data. For example, if the first communication device needs to send the filtered K output data to the second communication device, the first communication device may generate a data index, which may be used to indicate the MK data or the K data, and is not specifically limited here.

[0243] As an example, such as Figure 4 As shown, in step S401, the first communication device receives a training request from the second communication device and instructs the second communication device on the training configuration according to the training request.

[0244] S402, The second communication device acquires the first input data according to the training configuration;

[0245] S403, The second communication device sends the first input data to the first communication device;

[0246] S404. The first communication device processes the first input data using N models to obtain M output data. A monitor M0 deployed on the first communication device monitors at least one of the M output data and determines whether at least one of the M output data satisfies a first condition. Steps S401 to S404 correspond to... Figure 3 Step 301 in the illustrated embodiment. If not satisfied, the first communication device is instructed to execute step S406;

[0247] S405. The first communication device filters the M output data to obtain a data index and K output data, wherein step S405 corresponds to Figure 3 Step 302 in the illustrated embodiment;

[0248] S406. The first communication device sends a data index and K output data to the second communication device, wherein the K output data are used by the second communication device to execute step S407.

[0249] S407. The second communication device generates tag data corresponding to the K output data based on the K output data. This tag data is used by the second communication device to execute step S408;

[0250] S408. The second communication device trains an evaluation model based on the label data corresponding to the K output data.

[0251] Method 2: The K output data are obtained by reprocessing the first input data through N models. For example, the first communication device can re-input the first input data into N models based on the fifth information to obtain K output data. Thus, the first communication device can obtain high-quality and diverse inference results by regenerating the output data.

[0252] Optionally, the number of output data obtained by reprocessing the first input data through N models is the same as the number of output data obtained by processing the first input data through N models, i.e., M = K.

[0253] Method 3: The K output data are obtained by processing the first input data through P models, where the P models are different from the N models. Alternatively, the first communication device manages the N models to obtain P models, then reprocesses the first input data to obtain the K output data.

[0254] Optionally, the first communication device may instruct one or more communication devices to send first information, such that the recipient of the first information, i.e., the first communication device, can determine P models based on the first information. The first information includes model information, enabling the first communication device to manage N models based on this model information.

[0255] Optionally, the model information involved in this application may indicate one or more of the following: model identifier (or index), model parameters, model structure, sampling parameters, and number of samples. For example, the model parameters may include one or more of the following: model hyperparameters and model capability levels.

[0256] For example, the sampling parameters may include at least one of the following:

[0257] Temperature coefficient: A coefficient that controls probability normalization. For example, higher temperatures make the output more random, while lower temperatures make the output more deterministic.

[0258] Random coefficients: random sampling with a certain probability;

[0259] Top-K: Sample from the K outputs with the highest probability;

[0260] Top-P: Samples from the set of outputs with a cumulative probability greater than or equal to p.

[0261] Optionally, the model management involved in this application may include one or more of the following: model scheduling, model updating, model switching, or function rollback. The P models may comprise some or all of the N models, or the P models may be completely different from the N models.

[0262] As an example, such as Figure 5 As shown, steps S501 to S504 are the same as those described above. Figure 4 Steps S401 to S404 in the example shown are similar and will not be repeated here. Steps S501 to S504 correspond to... Figure 3 Step 301 in the illustrated embodiment.

[0263] S505, the first communication device switches N models to P models. The first communication device uses the P models to process the first input data and obtain K output data, wherein step S505 corresponds to... Figure 3 Step 302 in the illustrated embodiment;

[0264] S506, The first communication device sends the K output data to the second communication device;

[0265] Steps S507 to S508 are the same as those described above. Figure 4 Steps S407 to S408 in the example shown are similar and will not be described in detail here.

[0266] Method 4: The K output data are obtained by processing the second input data through N models, where the second input data is different from the first input data. Specifically, the first communication device acquires the second input data and processes it into N models to obtain K output data.

[0267] Optionally, the first communication device sends configuration information for the second input data, enabling the recipient of the configuration information (e.g., the second communication device) to determine the second input data based on the configuration information. For example, the configuration information may include a pre-trained model ID, the number of samples, or a pattern of the inference results, etc.

[0268] Optionally, the first communication device receives second information, which is used by the sender of the second information (e.g., the second communication device) to request configuration information for the second input data from the first communication device. Thus, the first communication device can send the configuration information for the second input data based on the second information.

[0269] As an example, such as Figure 6 As shown, steps S601 to S604 are the same as those described above. Figure 4 Steps S401 to S404 in the example shown are similar and will not be repeated here. Steps S601 to S604 correspond to... Figure 3 Step 301 in the illustrated embodiment.

[0270] S605. The first communication device sends an instruction message to the second communication device, which instructs the second communication device to restart model training.

[0271] S606, The second communication device sends a second training request to the first communication device, wherein the second training request is used to request configuration information of the second input data;

[0272] S607. In response to the second training request, the first communication device sends the second training configuration, i.e., the configuration information of the second input data, to the second communication device.

[0273] S608. The second communication device obtains the second input data according to the configuration information and sends the second input data to the first communication device.

[0274] S609. The first communication device processes the second input data using N models to obtain K output data, wherein steps S605 to S609 correspond to... Figure 3 Step 302 in the illustrated embodiment;

[0275] Steps S610 to S612 are the same as those described above. Figure 5 Steps S506 to S508 in the example shown are similar and will not be described in detail here.

[0276] Method 5: The K output data are obtained by processing the second input data through P models, where the second input data is different from the first input data, and the P models are different from the N models. Alternatively, the first communication device obtains K data by processing the updated input data through P models. The P models and the second input data can be referenced in the above embodiments, and will not be elaborated further here.

[0277] As an example, such as Figure 7 As shown, steps S701 to S708 are the same as those described above. Figure 6 Steps S601 to S608 in the example shown are similar and will not be repeated here. Steps S701 to S704 correspond to... Figure 3 Step 301 in the illustrated embodiment.

[0278] S709. The first communication device switches N models to P models. The first communication device uses the P models to process the second input data and obtain K output data; wherein, steps S705 to S709 correspond to Figure 3 Step 302 in the illustrated embodiment.

[0279] Steps S710 to S712 are the same as those described above. Figure 5 Steps S506 to S508 in the example shown are similar and will not be described in detail here.

[0280] Method 6: The first communication device can trigger the M0 event, thereby obtaining K output data based on M output data. The operation performed by the first communication device after triggering the M0 event can be one or more of the above-mentioned possible implementations, and is not specifically limited here.

[0281] It should be noted that the K output data determined by the first communication device can be used to generate label data, which is used to train the evaluation model; or it can be used as input data for the evaluation model to obtain inference results. The specific method is not limited here.

[0282] Please see Figure 8 This is another schematic diagram of the communication method provided in this application, which includes the following steps.

[0283] 801. The first communication device acquires X output data, wherein the X output data are obtained by processing the third input data through Y models, or the X output data are label data generated by the third input data, where X is a positive integer and Y is a positive integer less than or equal to X.

[0284] In one possible implementation, the third input data can be K output data. The details of these K output data can be found in the above embodiments, and will not be elaborated here.

[0285] Optionally, the Y models are used to determine the target output corresponding to the K output data; wherein the model processing performed by the Y models may include evaluation, sorting, optimization, screening, selection, or filtering. Accordingly, the Y models may be called evaluation models, personalized models, personalized evaluation models, reward models, or scoring models, etc.

[0286] Optionally, the label data is used to train Y models, which are then used to determine the target output corresponding to the K output data. In other words, the Y models are used to determine a personalized target output; that is, the Y models can be personalized models, and the aforementioned label data can be used to determine this personalized model. Therefore, the label data determined by the first communication device using the K output data can be used to determine a personalized model, which can then provide a personalized target output to improve the user experience.

[0287] In this embodiment, the method for obtaining the X output data can be referred to the foregoing. Figure 3 The method for obtaining the M output data in the illustrated embodiment will not be described in detail here.

[0288] It should be noted that when the evaluation model is not yet determined, these X output data are labeled data; after the evaluation model is determined, these X output data are the inference results of the evaluation model.

[0289] 802. The first communication device determines Z output data based on X output data; where Z is a positive integer. At least one of the Z output data is different from any one of the X output data.

[0290] In one possible implementation, the first communication device may determine the Z output data when at least one of the X output data does not satisfy the second condition.

[0291] Optionally, if at least one output data does not satisfy the second condition, it can be understood that the third parameter of at least one of the X output data satisfies the second preset condition, wherein the third parameter is used to indicate at least one of the distribution function, extreme value, mean, variance or interval of the output data.

[0292] Optionally, if at least one output data does not satisfy the second condition, it can be understood that the distribution function of the at least one output data is different from the preset distribution function.

[0293] Optionally, if at least one output data does not satisfy the second condition, it can be understood that the extreme value of the at least one output data is greater than the preset extreme value; or, the extreme value of the at least one output data is less than the preset extreme value.

[0294] Optionally, if at least one output data does not satisfy the second condition, it can be understood that the mean of the at least one output data is greater than the preset mean; or, the mean of the at least one output data is less than the preset mean.

[0295] Optionally, if at least one output data does not satisfy the second condition, it can be understood that the variance of the at least one output data is greater than the preset variance; or, the variance of the at least one output data is less than the preset variance.

[0296] Optionally, at least one output data does not satisfy the second condition, which can be understood as the interval of the at least one output data being different from the preset distribution interval.

[0297] As an example, the first communication device can judge at least one of the X output data and determine K output data.

[0298] As an example, the at least one output data can be determined by a monitor deployed on the first communication device. Specifically, the monitor can be a hardware device or a software device deployed on the first communication device, and no specific limitation is made here. In this embodiment, the monitor used to monitor the X output data is referred to as the M1 monitor.

[0299] In one possible implementation, the X output data points are labeled data, meaning they are used to train the evaluation model. Since the evaluation model is not yet determined during training, the M1 monitor is used to monitor the distribution of the ground truth (label values).

[0300] In another possible implementation, the X output data are the inference results of the evaluation model.

[0301] Optionally, when the M1 monitor determines that at least one of the X output data does not meet the second condition, the M1 monitor sends a sixth message to the first communication device, the sixth message indicating that at least one of the X output data does not meet the second condition, or the sixth message indicating that at least one of the X output data does not meet the second condition.

[0302] In this embodiment of the application, by determining Z output data that satisfy the second condition based on X output data during model training or model inference, the system can maintain the stability of its output or performance when subjected to disturbances or input changes, thereby improving the robustness of model training or model inference.

[0303] The first communication device can obtain Z output data in various ways, which will be explained below.

[0304] Method 1: The first communication device filters out the output data that does not meet the second condition from the X output data, obtaining Z output data. That is, the first communication device filters XZ output data. Alternatively, it can be said that the Z output data are contained in the X output data; among the X output data, the Z output data satisfy the second condition.

[0305] As an example, such as Figure 9 As shown, in step S901, the first communication device receives third input data from the second communication device;

[0306] S902. The first communication device generates tag data corresponding to the third input data based on the third input data, obtaining X output data. The M1 monitor deployed on the first communication device monitors at least one of the M output data and determines whether at least one of the X output data satisfies the second condition. If not, the first communication device is instructed to execute step S903; wherein, steps S901 to S902 correspond to... Figure 8 Step 801 in the illustrated embodiment;

[0307] S903. The first communication device filters the X output data to obtain Z output data, which are then used by the first communication device to execute step S904; wherein, step S903 corresponds to Figure 8 Step 802 in the illustrated embodiment;

[0308] S904. The first communication device trains an evaluation model based on the Z output data.

[0309] As an example, such as Figure 10 As shown, in step S1001, the first communication device receives third input data from the second communication device;

[0310] S1002. The first communication device uses Y models to infer the third input data and obtains X output data. The M1 monitor deployed on the first communication device monitors at least one of the M output data and determines whether at least one of the X output data satisfies the second condition. If not, the first communication device is instructed to execute step S1003; wherein steps S1001 to S1002 correspond to... Figure 8 Step 801 in the illustrated embodiment;

[0311] S1003: The first communication device filters the X output data to obtain Z output data. These Z output data represent the inference results of the evaluation model. Step S1003 corresponds to... Figure 8 Step 802 in the illustrated embodiment.

[0312] Method 2: The first communication device regenerates the label data corresponding to the third input data, or the first communication device reprocesses the third input data using Y models to obtain Z output data. For example, the first communication device can re-input the third input data into Y models based on the sixth information to obtain Z output data. Thus, the first communication device can obtain high-quality and diverse inference results by regenerating the output data.

[0313] Optionally, the number of output data obtained by reprocessing the first input data through Y models is the same as the number of output data obtained by processing the first input data through Y models, i.e., X = Z.

[0314] Method 3: The Z output data are obtained by processing the third input data through Q models, where the Q models are different from the Y models. Alternatively, the first communication device manages the Y models to obtain Q models, then reprocesses the third input data to obtain the Z output data.

[0315] Method 4: Z output data are obtained by processing X output data from the third input data using different label generation methods. Here, different label generation methods can be understood as switching the algorithm, model, or personnel used to generate the labels.

[0316] Optionally, the first communication device may instruct one or more communication devices to send fourth information, such that the recipient of the fourth information, i.e., the first communication device, can determine Q models based on the fourth information. The fourth information includes model information, enabling the first communication device to manage Y models based on this model information.

[0317] Optionally, the first communication device may instruct one or more communication devices to send fourth information, such that the recipient of the fourth information, i.e., the first communication device, can determine Q models based on the fourth information. The fourth information includes model information, enabling the first communication device to manage Y models based on this model information.

[0318] Optionally, the model information involved in this application may indicate one or more of the following: model identifier (or index), model parameters, model structure, sampling parameters, and number of samples. For example, the model parameters may include one or more of the following: model hyperparameters and model capability levels.

[0319] For example, the sampling parameters may include at least one of the following:

[0320] Temperature coefficient: A coefficient that controls probability normalization. For example, higher temperatures make the output more random, while lower temperatures make the output more deterministic.

[0321] Random coefficients: random sampling with a certain probability;

[0322] Top-K: Sample from the K outputs with the highest probability;

[0323] Top-P: Samples from the set of outputs with a cumulative probability greater than or equal to p.

[0324] Optionally, the model management involved in this application may include one or more of the following: model scheduling, model updating, model switching, or function rollback.

[0325] As an example, such as Figure 11 As shown, in step S1101, the first communication device receives third input data from the second communication device;

[0326] S1102. The first communication device uses Y models to infer the third input data and obtains X output data. The M1 monitor deployed on the first communication device monitors at least one of the M output data and determines whether at least one of the X output data satisfies the second condition. If not, the first communication device is instructed to execute step S1103; wherein steps S1101 to S1102 correspond to... Figure 8 Step 801 in the illustrated embodiment;

[0327] S1103, the first communication device switches the Y models to Q models. The first communication device uses the Q models to process the third input data and obtain Z output data. Step S1103 corresponds to... Figure 8 Step 802 in the illustrated embodiment.

[0328] Optionally, the first communication device can send a fourth message, enabling the recipient of the fourth message, such as the second communication device, to retrain the evaluation model based on the fourth message, thereby obtaining the Q models. The second communication device sends the Q models to the first communication device, so that the first communication device can use the Q models to process the third input data and obtain Z output data.

[0329] In another possible implementation, the Z output data are obtained by processing the fourth input data through Y models, or the Z output data are the label data corresponding to the fourth input data, wherein the fourth input data is different from the third input data. Specifically, the first communication device acquires the fourth input data and processes it into Y models to obtain Z output data, or generates labels corresponding to the fourth input data to obtain Z output data.

[0330] In another possible implementation, the first communication device can trigger the M1 event to obtain Z output data based on X output data. The operation performed by the first communication device after triggering the M1 event can be one or more of the above-mentioned possible implementations, and is not specifically limited here.

[0331] In another possible implementation, the first communication device can report an M1 event, enabling the recipient of the M1 event, such as the second communication device, to make a decision based on the M1 event. For example, the second communication device can instruct the first communication device to send a new training request based on the M1 event to retrain and obtain the fourth input data. Alternatively, the second communication device can switch the model deployed by the second communication device based on the M1 event, thereby retraining to obtain the fourth input data.

[0332] In this embodiment, the method by which the second communication device retrains to obtain the fourth input data can refer to the method in the above embodiment where the first communication device determines K output data based on M output data, and the specifics will not be repeated here.

[0333] As an example, such as Figure 12 As shown, in S1201, the first communication device receives third input data from the second communication device;

[0334] S1202: The first communication device uses Y models to infer the third input data and obtains X output data. The M1 monitor deployed on the first communication device monitors at least one of the M output data and determines whether at least one of the X output data satisfies the second condition. If not, the first communication device is instructed to execute step S1203; wherein steps S1201 to S1202 correspond to... Figure 8 Step 801 in the illustrated embodiment.

[0335] S1203. The first communication device sends an M1 event to the second communication device, causing the second communication device to retrain and obtain the fourth input data based on the M1 event.

[0336] S1204, The second communication device sends the fourth input data to the first communication device;

[0337] S1205. The first communication device generates tag data corresponding to the fourth input data, obtaining Z output data. The Z output data are used by the first communication device to execute step S1206; wherein, steps S1203 to S1205 correspond to Figure 8 Step 802 in the illustrated embodiment;

[0338] S1206. The first communication device trains an evaluation model based on the Z output data.

[0339] As an example, such as Figure 13 As shown, in S1301, the first communication device receives third input data from the second communication device;

[0340] S1302. The first communication device uses Y models to infer the third input data and obtains X output data. The M1 monitor deployed on the first communication device monitors at least one of the M output data and determines whether at least one of the X output data satisfies the second condition. If not, the first communication device is instructed to execute step S1303; wherein steps S1301 to S1302 correspond to... Figure 8 Step 801 in the illustrated embodiment;

[0341] S1303, The first communication device sends an M1 event to the second communication device, and the second communication device retrains to obtain the fourth input data based on the M1 event;

[0342] S1304, The second communication device sends the fourth input data to the first communication device;

[0343] S1305, the first communication device processes the fourth input data using Y models to obtain Z output data. Steps S1303 to S1305 correspond to... Figure 8 Step 802 in the illustrated embodiment.

[0344] It should be understood that, Figure 8 The embodiments shown are the same as those described above. Figure 3 The illustrated embodiments can be used in combination. Alternatively, Figure 8 The embodiments shown are the same as those described above. Figure 3 The embodiments shown are not combined but applied independently, and no specific limitation is made here. For example, the M0 monitor and the M1 monitor can be deployed on the same communication device or on different communication devices, as illustrated below.

[0345] Example 1, such as Figure 14 As shown, the M0 monitor is deployed on the first communication device, and the M1 monitor is deployed on the second communication device.

[0346] S1401, The first communication device receives a training request from the second communication device, the training request being used by the second communication device to train an evaluation model;

[0347] S1402. The first communication device sends the training configuration of the first input data to the second communication device according to the training request;

[0348] S1403. The second communication device collects the first input data based on the training configuration of the first input data and sends the first input data to the first communication device.

[0349] S1404. The first communication device processes the first input data using N models to obtain M output data. The M0 monitor monitors the M output data and, based on the M output data, instructs the first communication device to determine K output data. The method for determining the K output data can be referred to the foregoing. Figure 3 The specific implementation examples shown will not be described in detail here;

[0350] S1405. The first communication device sends the K output data to the second communication device. The K output data are used by the second communication device to execute step S1406.

[0351] S1406. The second communication device generates tag data corresponding to the K output data based on the K output data, obtaining X output data. The M1 monitor monitors the X output data and instructs the second communication device to determine Z output data based on the X output data. The Z output data is used by the second communication device to execute step S1407.

[0352] S1407. The second communication device trains the evaluation model based on the Z output data.

[0353] Example 2, such as Figure 15As shown, both the M0 and M1 monitors are deployed on the second communication device. Steps S1501 to S1503 are the same as described above. Figure 14 Steps S1401 to S1403 in the example shown are similar and will not be described in detail here.

[0354] S1504. The first communication device uses N models to process the first input data and obtains M output data.

[0355] S1505, The first communication device sends M output data to the second communication device. The M0 monitor deployed on the second communication device monitors the M output data and instructs the second communication device to determine K output data based on the M output data;

[0356] S1506. The second communication device uses Y models to infer the K output data to obtain X output data. The M1 monitor monitors the X output data and instructs the second communication device to determine Z output data based on the X output data.

[0357] In this embodiment, by monitoring the output of the model, abnormal states that occur during training or inference can be detected in a timely manner, and new output data can be obtained. This solves the problem of abnormal states that occur during training or inference, thereby improving the robustness of model training or model inference.

[0358] The information transmission method in the embodiments of this application has been described above. The communication device in the embodiments of this application is described below. Please refer to [link / reference]. Figure 16 The communication device 1600 can be used to perform Figure 3 or Figure 8 The process executed by the first or second communication device in the illustrated embodiments can be specifically referred to the relevant descriptions in the foregoing method embodiments. The communication device 1600 can be a network device, a component or device applied to a network device (e.g., a processor, chip, or chip system), or a logic module or software capable of implementing all or part of the functions of a network device. The communication device can also be a terminal device, a component or device applied to a terminal device (e.g., a processor, chip, or chip system), or a logic module or software capable of implementing all or part of the functions of a terminal device.

[0359] The communication device 1600 includes an interface module 1601 and a processing module 1602.

[0360] The processing module 1602 is used for data processing. The interface module 1601 can implement corresponding communication functions. The interface module 1601 can also be called a communication interface or a communication module.

[0361] Optionally, the communication device 1600 may further include a storage module, which can be used to store program code, program instructions and / or data. The processing module 1602 can read the instructions and / or data in the storage module so that the communication device 1600 can implement the aforementioned method embodiments.

[0362] The communication device 1600 can be used to perform the actions performed by the first or second communication device in the method embodiments described above. For example, it can be the first communication device, a communication module within the first communication device, or a circuit or chip responsible for communication functions within the first communication device. The communication device 1600 can be the first communication device or a component configurable within the first communication device. The processing module 1602 is used to perform processing-related operations on the side of the first or second communication device in the method embodiments described above. The interface module 1601 is used to perform receiving-related operations on the side of the first or second communication device in the method embodiments described above.

[0363] Optionally, interface module 1601 may include a sending module and a receiving module. The sending module is used to perform the sending operation in the above method embodiments. The receiving module is used to perform the receiving operation in the above method embodiments.

[0364] It should be noted that the communication device 1600 may include a transmitting module but not a receiving module. Alternatively, the communication device 1600 may include a receiving module but not a transmitting module. Specifically, it depends on whether the above-described scheme performed by the communication device 1600 includes both transmitting and receiving actions. For example, the communication device 1600 is used to perform the above-described... Figure 3 or Figure 8 The actions performed by the first or second communication device in the illustrated embodiments are described above. For details, please refer to the above. Figure 3 or Figure 8 The relevant descriptions in the illustrated embodiments will not be elaborated here.

[0365] For example, the communication device 1600 is used to execute the following scheme:

[0366] The processing module 1602 is used to acquire M output data. The M output data are obtained by processing the first input data through N models. Any one of the N models is used to perform inference on the first input data. M is a positive integer and N is a positive integer less than or equal to M.

[0367] The processing module 1602 is also used to determine K output data based on M output data, where K is a positive integer.

[0368] In one possible implementation, the processing module 1602 is used to acquire M output data, including: the processing module 1602 receives the M output data through the interface module 1601.

[0369] In another possible implementation, the processing module 1602 is further configured to determine K output data based on M output data, including: the processing module 1602 is specifically configured to determine K output data when at least one of the M output data does not satisfy a first condition.

[0370] In another possible implementation, interface module 1601 is also used to receive first information, which is used to determine P models.

[0371] In another possible implementation, the interface module 1601 is also used to receive second information, which is configuration information for requesting second input data;

[0372] Interface module 1601 is also used to send configuration information.

[0373] In another possible implementation, interface module 1601 is also used to send third information, which indicates K output data.

[0374] In another possible implementation, interface module 1601 is also used to receive fourth information, which is used to determine Q models, any one of the Q models being used to score K data.

[0375] In another possible implementation, interface module 1601 is also used to obtain X output data, which are obtained by processing K output data through Y models, or, the X output data are label data generated from K output data, and any one of the Y models is used to score the K output data, where X is a positive integer and Y is a positive integer less than or equal to X.

[0376] The processing module 1602 is also used to determine Z output data based on X output data, where Z output data is obtained by processing X output data and Z is a positive integer.

[0377] In another possible implementation, the processing module 1602 is further configured to determine Z output data based on X output data, including: the processing module 1602 is specifically configured to determine Z output data when at least one of the X output data does not satisfy the second condition.

[0378] For example, the communication device 1600 is used to execute the following scheme:

[0379] The processing module 1602 is used to acquire X output data, which are obtained by processing the third input data through Y models, or the X output data are label data generated from the third input data. Any one of the Y models is used to score the third input data, where X is a positive integer and Y is a positive integer less than or equal to X.

[0380] The processing module 1602 is also used to determine Z output data based on X output data, where Z is a positive integer.

[0381] In one possible implementation, the processing module 1602 is used to acquire X output data, including: the processing module 1602 receives the X output data through the interface module 1601.

[0382] In another possible implementation, the processing module 1602 is further configured to determine Z output data based on X output data, including: the processing module 1602 is specifically configured to determine Z output data when at least one of the X output data does not satisfy the second condition.

[0383] In another possible implementation, interface module 1601 is also used to receive third information, which is used to indicate third input data.

[0384] In another possible implementation, the interface module 1601 is also used to send first information, which is used to determine P models, and the P models are used to reason about the first input data to obtain the third input data, where P is a positive integer.

[0385] In another possible implementation, the interface module 1601 is also used to send second information, which is used to request configuration information for the second input data. The second input data is used to process the second input data through N models to obtain the third input data. Any one of the N models is used to infer the second input data, where N is a positive integer.

[0386] Receive configuration information.

[0387] In another possible implementation, interface module 1601 is specifically used to send fourth information, which is used to determine Q models.

[0388] Please see Figure 17 This is another schematic structural diagram of the communication device 1700 provided in this application. The communication device 1700 includes a logic circuit 1701 and an input / output interface 1702. The communication device 1700 can be a chip or an integrated circuit.

[0389] in, Figure 16 The interface module 1601 shown can be a communication interface, which can be... Figure 17 The input / output interface 1702 may include an input interface and an output interface. Alternatively, the communication interface may also be a transceiver circuit, which may include an input interface circuit and an output interface circuit.

[0390] Optionally, logic circuit 1701 is used to acquire M output data, which are obtained by processing the first input data through N models. Any one of the N models is used to infer the first input data, where M is a positive integer and N is a positive integer less than or equal to M. Logic circuit 1701 is also used to determine K output data based on the M output data, where K is a positive integer.

[0391] Optionally, logic circuit 1701 is used to acquire X output data, which are obtained by processing the third input data through Y models, or the X output data are label data generated from the third input data, and any one of the Y models is used to score the third input data, where X is a positive integer and Y is a positive integer less than or equal to X; logic circuit 1701 is also used to determine Z output data based on the X output data, where Z is a positive integer.

[0392] The logic circuit 1701 and the input / output interface 1702 can also perform other steps performed by the first or second communication device in any embodiment and achieve corresponding beneficial effects, which will not be elaborated here.

[0393] In one possible implementation, Figure 16 The processing module 1602 shown can be Figure 17 The logic circuit 1701 in the middle.

[0394] Optionally, the logic circuit 1701 can be a processing device, the functions of which can be partially or entirely implemented in software.

[0395] Optionally, the processing apparatus may include a memory and a processor, wherein the memory is used to store a computer program, and the processor reads and executes the computer program stored in the memory to perform the corresponding processing and / or steps in any of the method embodiments.

[0396] Optionally, the processing device may consist of only a processor. A memory for storing computer programs is located outside the processing device, and the processor is connected to the memory via circuitry / wires to read and execute the computer programs stored in the memory. The memory and processor may be integrated together or physically independent of each other.

[0397] Optionally, the processing device may be one or more chips, or one or more integrated circuits. For example, the processing device may be one or more field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), system on-chips (SoCs), central processors (CPUs), network processors (NPs), digital signal processors (DSPs), microcontroller units (MCUs), programmable logic devices (PLDs), or other integrated chips, or any combination of the above chips or processors.

[0398] The following describes a communication device provided in an embodiment of this application. Please refer to [link / reference]. Figure 18 , Figure 18 This is a schematic diagram of a communication device provided in an embodiment of this application. The communication device may be the first communication device or the second communication device in the above method embodiments, or it may be a chip, chip system, or processor that supports the first communication device or the second communication device in implementing the above method. This communication device can be used to implement the methods described in the above method embodiments, and for details, please refer to the description in the above method embodiments.

[0399] The communication device may include one or more processors 1801, which are connected to a memory 1802, an input / output unit 1803, and a bus 1804. The processor 1801 may be a general-purpose processor or a dedicated processor, such as a baseband processor or a central processing unit (CPU). The baseband processor can be used to process communication protocols and communication data, while the CPU can be used to control the communication device (e.g., base station, baseband chip, terminal, terminal chip, DU or CU, etc.), execute software programs, and process data from the software programs.

[0400] Optionally, the communication device may include one or more memories 1802, which may store instructions that can be executed on the processor 1801 to cause the communication device to perform the methods described in the above method embodiments. Optionally, the memories 1802 may also store data. The processor 1801 and the memories 1802 may be configured separately or integrated together.

[0401] Optionally, the communication device may also include a transceiver and an antenna. A transceiver, also called a transceiver unit, transceiver, or transceiver circuit, is used to implement transmission and reception functions. A transceiver may include a receiver and a transmitter; the receiver, also called a receiver circuit, is used to implement the receiving function; the transmitter, also called a transmitter or transmitting circuit, is used to implement the transmitting function.

[0402] In another possible design, the processor 1801 may include a transceiver for implementing receive and transmit functions. For example, the transceiver may be a transceiver circuit, an interface, or an interface circuit. The transceiver circuit, interface, or interface circuit for implementing receive and transmit functions may be separate or integrated. The aforementioned transceiver circuit, interface, or interface circuit may be used for reading and writing code / data, or it may be used for transmitting or relaying signals.

[0403] In another possible design, the processor 1801 may optionally store instructions that, when executed, cause the communication device to perform the methods described in the above method embodiments. The instructions may be stored in the processor 1801; in this case, the processor 1801 may be implemented in hardware.

[0404] In another possible design, the communication device may include a circuit that can perform the transmitting or receiving or communication functions of the first or second communication device in the aforementioned method embodiments. The processor and transceiver described in this application embodiment can be implemented on integrated circuits (ICs), analog ICs, radio frequency integrated circuits (RFICs), mixed-signal ICs, application-specific integrated circuits (ASICs), printed circuit boards (PCBs), electronic devices, etc. The processor and transceiver can also be manufactured using various IC process technologies, such as complementary metal oxide semiconductors (CMOS), n-type metal-oxide-semiconductor (NMOS), p-type metal oxide semiconductors (PMOS), bipolar junction transistors (BJTs), bipolar CMOS (BiCMOS), silicon germanium (SiGe), gallium arsenide (GaAs), etc.

[0405] The communication device described in the above embodiments may be a first communication device or a second communication device, but the scope of the communication device described in the embodiments of this application is not limited thereto, and the structure of the communication device may vary. Figure 18 The communication device can be a standalone device or part of a larger device. For example, the communication device can be:

[0406] (1) Independent integrated circuit IC, or chip, or chip system or subsystem;

[0407] (2) A collection of one or more ICs, optionally including a storage component for storing data and instructions;

[0408] (3) ASIC, such as modem;

[0409] (4) Modules that can be embedded in other devices;

[0410] (5) Receivers, terminals, smart terminals, cellular phones, wireless devices, handheld devices, mobile units, vehicle-mounted devices, network devices, cloud devices, artificial intelligence devices, etc.

[0411] (6) Others, etc.

[0412] Please see Figure 19 The above-described embodiments of the communication device 1900, provided as an example of the present application, are structural schematic diagrams. Specifically, the communication device 1900 can be a network device as described in the above embodiments. Figure 19 The communication device illustrated is implemented through a network device (or a component within a network device), the structure of which can be referenced. Figure 19 The structure shown.

[0413] The communication device 1900 includes at least one processor 1911 and at least one network interface 1914. Optionally, the communication device further includes at least one memory 1912, at least one transceiver 1913, and one or more antennas 1915. The processor 1911, memory 1912, transceiver 1913, and network interface 1914 are connected, for example, via a bus. In this embodiment, the connection may include various interfaces, transmission lines, or buses, etc., and this embodiment is not limited thereto. The antenna 1915 is connected to the transceiver 1913. The network interface 1914 enables the communication device to communicate with other communication devices through a communication link. For example, the network interface 1914 may include a network interface between the communication device and core network equipment, such as an S1 interface; the network interface may also include a network interface between the communication device and other communication devices (e.g., other network devices or core network equipment), such as an X2 or Xn interface.

[0414] in, Figure 16 The interface module 1601 shown can be a communication interface, which can be... Figure 19 The network interface 1914 may include an input interface and an output interface. Alternatively, the network interface 1914 may also be a transceiver circuit, which may include input interface circuitry and output interface circuitry.

[0415] The processor 1911 is primarily used to process communication protocols and communication data, control the entire communication device, execute software programs, and process data from the software programs, for example, to support the communication device in performing the actions described in the embodiments. The communication device may include a baseband processor and a central processing unit (CPU). The baseband processor is primarily used to process communication protocols and communication data, while the CPU is primarily used to control the entire terminal device, execute software programs, and process data from the software programs. Figure 19 The processor 1911 can integrate the functions of a baseband processor and a central processing unit. Those skilled in the art will understand that the baseband processor and the central processing unit can also be independent processors interconnected via technologies such as buses. Those skilled in the art will understand that a terminal device can include multiple baseband processors to adapt to different network standards, and a terminal device can include multiple central processing units to enhance its processing capabilities. The various components of the terminal device can be connected via various buses. The baseband processor can also be described as a baseband processing circuit or a baseband processing chip. The central processing unit can also be described as a central processing circuit or a central processing chip. The function of processing communication protocols and communication data can be built into the processor or stored in memory as a software program, with the processor executing the software program to implement the baseband processing function.

[0416] The memory is primarily used to store software programs and data. The memory 1912 can exist independently or be connected to the processor 1911. Optionally, the memory 1912 can be integrated with the processor 1911, for example, integrated into a single chip. The memory 1912 can store program code that executes the technical solutions of the embodiments of this application, and its execution is controlled by the processor 1911. The various types of computer program code being executed can also be considered as drivers for the processor 1911.

[0417] Figure 19 Only one memory and one processor are shown. In actual terminal devices, there may be multiple processors and multiple memories. Memory can also be called storage medium or storage device, etc. Memory can be a storage element on the same chip as the processor, i.e., an on-chip storage element, or it can be a separate storage element; this application does not limit this.

[0418] Transceiver 1913 can be used to support the reception or transmission of radio frequency (RF) signals between a communication device and a terminal. Transceiver 1913 can be connected to antenna 1915. Transceiver 1913 includes a transmitter Tx and a receiver Rx. Specifically, one or more antennas 1915 can receive RF signals. The receiver Rx of transceiver 1913 is used to receive the RF signals from the antennas, convert the RF signals into digital baseband signals or digital intermediate frequency (IF) signals, and provide the digital baseband signals or IF signals to processor 1911 so that processor 1911 can perform further processing on the digital baseband signals or IF signals, such as demodulation and decoding. In addition, the transmitter Tx in transceiver 1913 is also used to receive modulated digital baseband signals or IF signals from processor 1911, convert the modulated digital baseband signals or IF signals into RF signals, and transmit the RF signals through one or more antennas 1915. Specifically, the receiver Rx can selectively perform one or more stages of downmixing and analog-to-digital conversion on the radio frequency signal to obtain a digital baseband signal or a digital intermediate frequency (IF) signal. The order of these downmixing and IF conversion processes is adjustable. The transmitter Tx can selectively perform one or more stages of upmixing and digital-to-analog conversion on the modulated digital baseband signal or digital IF signal to obtain a radio frequency signal. The order of these upmixing and IF conversion processes is also adjustable. The digital baseband signal and the digital IF signal can be collectively referred to as digital signals.

[0419] The transceiver 1913 can also be called a transceiver unit, transceiver, transceiver device, etc. Optionally, the device in the transceiver unit that performs the receiving function can be regarded as the receiving unit, and the device in the transceiver unit that performs the transmitting function can be regarded as the transmitting unit. That is, the transceiver unit includes a receiving unit and a transmitting unit. The receiving unit can also be called a receiver, input port, receiving circuit, etc., and the transmitting unit can be called a transmitter, transmitter, or transmitting circuit, etc.

[0420] It should be noted that, Figure 19 The communication device 1900 shown can be used to implement the steps implemented by the network device in the aforementioned method embodiments, and to achieve the corresponding technical effects of the network device. Figure 19 The specific implementation of the communication device 1900 shown can be referred to the description in the foregoing method embodiments, and will not be repeated here.

[0421] Please see Figure 20 The above-described embodiments of the communication device provided in this application are schematic diagrams of the structure of the communication device.

[0422] It is understood that the communication device 100 includes, for example, modules, units, elements, circuits, or interfaces, which are appropriately configured together to execute the technical solutions provided in this application. The communication device 100 may be the terminal device or network device described above, or a component (e.g., a chip) within these devices, used to implement the methods described in the following method embodiments. The communication device 100 includes one or more processors 101. The processor 101 may be a general-purpose processor or a dedicated processor, for example, a baseband processor or a central processing unit. The baseband processor can be used to process communication protocols and communication data, and the central processing unit can be used to control the communication device (e.g., a RAN node, terminal, or chip), execute software programs, and process data from the software programs.

[0423] Optionally, in one design, processor 101 may include program 103 (sometimes also referred to as code or instructions), which can be executed on processor 101 to cause communication device 100 to perform the methods described in the embodiments below. In yet another possible design, communication device 100 includes circuitry (…). Figure 20 (Not shown).

[0424] Optionally, the communication device 100 may include one or more memories 102 storing a program 104 (sometimes referred to as code or instructions), which can be run on the processor 101 to cause the communication device 100 to perform the methods described in the above method embodiments.

[0425] Optionally, the processor 101 and / or memory 102 may include AI modules 107 and 108, which are used to implement AI-related functions. The AI ​​modules can be implemented through software, hardware, or a combination of both. For example, the AI ​​module may include a radio intelligence control (RIC) module. For example, the AI ​​module may be a near real-time RIC or a non-real-time RIC.

[0426] Optionally, the processor 101 and / or memory 102 may also store data. The processor and memory may be configured separately or integrated together.

[0427] Optionally, the communication device 100 may further include a transceiver 105 and / or an antenna 106. The processor 101, sometimes referred to as a processing unit, controls the communication device (e.g., a RAN node or terminal). The transceiver 105, sometimes referred to as a transceiver unit, transceiver, transceiver circuit, or transceiver, is used to realize the transmission and reception functions of the communication device through the antenna 106.

[0428] in, Figure 16 The processing module 1602 shown may be a processor 101. Figure 16 The interface module 1601 shown can be a communication interface, which can be... Figure 20 The transceiver 105 may include an input interface and an output interface. Alternatively, the transceiver 105 may also be a transceiver circuit, which may include an input interface circuit and an output interface circuit.

[0429] It should be understood that the processor in the embodiments of this application can be an integrated circuit chip with signal processing capabilities. In implementation, the steps of the above method embodiments can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor described above can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0430] It is understood that the memory in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAK are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM). It should be noted that the memory used in the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0431] This application also provides a computer-readable storage medium including instructions that, when executed on a computer, cause the computer to perform the methods described in the foregoing embodiments.

[0432] This application also provides a computer program product containing instructions that, when run on a computer, cause the computer to perform the methods described in the foregoing embodiments.

[0433] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0434] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0435] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0436] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0437] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0438] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. 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 via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., high-density digital video discs (DVDs)), or semiconductor media (e.g., solid-state disks (SSDs)).

Claims

1. A communication method, characterized in that, include: M output data are obtained, wherein the M output data are obtained by processing the first input data through N models, and any one of the N models is used to infer one or more output data from the first input data, wherein M is a positive integer and N is a positive integer less than or equal to M; Based on the M output data, K output data are determined, where K is a positive integer.

2. The method according to claim 1, characterized in that, The process of determining K output data based on the M output data includes: If at least one of the M output data does not satisfy the first condition, the K output data are determined.

3. The method according to claim 2, characterized in that, The at least one output data does not satisfy the first condition, including: The first parameter of the at least one output data is lower than a first preset value, and the first parameter is used to indicate the degree of correlation between the at least one output data and the first input data; and / or, The second parameter of the at least one output data satisfies the first preset condition, and the second parameter is used to indicate the diversity of the at least one output data.

4. The method according to any one of claims 1 to 3, characterized in that, The K output data satisfy any one of the following: The K output data are included in the M output data; wherein, among the M output data, the K output data satisfy a first condition; The K output data are obtained by reprocessing the first input data through the N models; The K output data are obtained by processing the first input data through P models, and the P models are different from the N models; The K output data are obtained by processing the second input data through the N models, and the second input data is different from the first input data; or, The K output data are obtained by processing the second input data through the P models.

5. The method according to claim 4, characterized in that, The method further includes: Receive first information, which is used to determine the P models.

6. The method according to claim 4, characterized in that, The method further includes: Receive second information, which is used to request configuration information for the second input data; Send the configuration information for the second input data.

7. The method according to any one of claims 1 to 6, characterized in that, The method further includes: Send a third message, which is used to indicate the K output data.

8. The method according to claim 7, characterized in that, The method further includes: Receive fourth information, which is used to determine Q models, and any one of the Q models is used to score the K data.

9. The method according to any one of claims 1 to 6, characterized in that, The method further includes: Z output data are determined based on the X output data, where Z is a positive integer; wherein the X output data are obtained by processing the K output data through Y models, or the X output data are label data generated from the K output data, and any one of the Y models is used to score the K output data, where X is a positive integer and Y is a positive integer less than or equal to X.

10. The method according to claim 9, characterized in that, The process of determining Z output data based on the X output data includes: If at least one of the X output data does not satisfy the second condition, the Z output data are determined.

11. The method according to claim 10, characterized in that, The at least one output data does not satisfy the second condition, including: The third parameter of at least one of the Z output data satisfies the second preset condition, and the third parameter is used to indicate at least one of the distribution function, extreme value, mean, variance or interval of the at least one output data.

12. The method according to any one of claims 9 to 11, characterized in that, The Z output data satisfy any one of the following: The Z output data are included in the X output data; wherein, among the X output data, the Z output data satisfy the second condition; The Z output data are obtained by reprocessing the K output data through the Y models; The Z output data are obtained by processing the K output data through Q models, where at least one of the Q models is different from the Y models, and Q is an integer greater than or equal to Y; The Z output data are obtained by processing the G output data through the Y models. At least one of the G output data is different from the K output data. The G output data are obtained by processing the third input data through the N models. G is an integer greater than or equal to Y. The Z output data are obtained by processing the G output data through the Q models; or... The Z output data are obtained by processing the G output data, the G output data are the tag data corresponding to the K output data, at least one of the G output data is different from the K output data, and G is an integer greater than or equal to Y.

13. A communication method, characterized in that, include: Obtain X output data, wherein the X output data are obtained by processing the fourth input data through Y models, or the X output data are label data generated from the fourth input data, and any one of the Y models is used to score the fourth input data, wherein X is a positive integer and Y is a positive integer less than or equal to X; Z output data are determined based on the X output data, where Z is a positive integer.

14. The method according to claim 13, characterized in that, The process of determining Z output data based on the X output data includes: If at least one of the X output data does not satisfy the second condition, the Z output data are determined.

15. The method according to claim 14, characterized in that, The at least one output data does not satisfy the second condition, including: At least one of the Z output data satisfies the second preset condition in its third parameter, which indicates at least one of the distribution function, extreme value, mean, variance, or interval of the output data.

16. The method according to any one of claims 13 to 15, characterized in that, The Z output data satisfy any one of the following: The Z output data are included in the X output data; wherein, among the X output data, the Z output data satisfy the second condition; The Z output data are obtained by reprocessing the fourth input data through the Y models; The Z output data are obtained by processing the fourth input data through Q models, where at least one of the Q models is different from the Y models, and Q is an integer greater than or equal to Y; The Z output data are obtained by processing the fifth input data through the Y models, and at least one of the output data in the fifth input data is different from the fourth input data; The Z output data are obtained by processing the fifth input data through the Q models; or, The Z output data are label data generated from the fifth input data.

17. The method according to any one of claims 13 to 16, characterized in that, The fourth input data includes K output data, where K is a positive integer, and the method further includes: Receive third information, which is used to indicate the K output data.

18. The method according to any one of claims 13 to 17, characterized in that, The method further includes: Send first information, which is used to determine P models, which are used to infer the fourth input data, where P is a positive integer.

19. The method according to any one of claims 13 to 18, characterized in that, The method further includes: Send a second message, which is used to request configuration information for the second input data. The second input data is used to process the second input data through N models to obtain the fourth input data. Any one of the N models is used to infer the second input data, where N is a positive integer. The configuration information received from the second input data.

20. The method according to any one of claims 13 to 19, characterized in that, The method further includes: Send a fourth message, which is used to determine the Q models.

21. A communication device, characterized in that, include: An interface module is used to acquire M output data, which are obtained by processing the first input data through N models. Any one of the N models is used to perform inference on the first input data. M is a positive integer and N is a positive integer less than or equal to M. The processing module is used to determine K output data based on the M output data, where K is a positive integer.

22. A communication device, characterized in that, include: An interface module is used to acquire X output data, wherein the X output data are obtained by processing the fourth input data through Y models, or the X output data are label data generated from the fourth input data, and any one of the Y models is used to score the fourth input data, wherein X is a positive integer and Y is a positive integer less than or equal to X; The processing module is used to determine Z output data based on the X output data, wherein the Z output data are obtained by processing the X output data, and Z is a positive integer.

23. A communication device, characterized in that, include: A processor for executing a program that causes the communication device to perform the method as described in any one of claims 1 to 12.

24. A communication device, characterized in that, include: A processor for executing a program that causes the communication device to perform the method as described in any one of claims 13 to 20.

25. A communication system, characterized in that, include: A communication device for performing any of the methods described in steps 1 to 12, and a communication device for performing any of the methods described in claims 13 to 20.

26. A computer-readable storage medium comprising instructions that, when executed on a computer, cause the computer to perform the method as claimed in any one of claims 1 to 12, or cause the computer to perform the method as claimed in any one of claims 13 to 20.

27. A computer program product comprising instructions that, when run on a computer, causes the computer to perform the method as claimed in any one of claims 1 to 12, or causes the computer to perform the method as claimed in any one of claims 13 to 20.

Images