Data processing method, training method and related device

Abstract

The invention provides a data processing method, which is applied to a wireless communication system with an artificial intelligence (AI) processing capability, and is applied to a wireless communication system with a wireless communication system with a wireless communication system with a wireless communication system with a wireless communication system with a wireless communication system with a wireless communication system with a wireless communication system with a wireless communication system with a wireless communication system. According to the method, machine learning models with the same structure are deployed on different communication devices, and a plurality of devices are combined to complete data processing, so that the quality of the obtained data is continuously improved, the data processing pressure of each device is reduced, and it is ensured that the device with weak computing power can also obtain the data with the required quality.

Description

A data processing method, training method and related device Technical Field

[0001] The present application relates to the field of artificial intelligence (AI) technology, and in particular to a data processing method, a training method, and related devices. Background Art

[0002] With the advent of the big data era, data volumes are growing rapidly. Extracting valuable information from this massive data has become a pressing issue in most application scenarios. Currently, machine learning models are often used to process data and extract valuable information from it.

[0003] Before using a machine learning model to process data, it is often necessary to train the model with a large amount of data to achieve high accuracy. However, in practical applications, high-quality data suitable for training machine learning models is often difficult to obtain. Especially in wireless communication networks, due to the influence of interference and noise, the acquired data often needs to undergo complex processing to obtain high-quality data. Therefore, how to obtain high-quality data for machine learning model training with low complexity is currently a hot research topic.

[0004] In related technologies, the process of processing raw data to obtain high-quality data can be considered a data denoising process. By applying multiple denoising operations to raw data using a machine learning model, high-quality data can be obtained. This entire process can be modeled using a Markov chain. However, the training process of the machine learning model used for denoising is highly complex, and using this machine learning model for inference, i.e., denoising the data, often requires performing a large number of denoising operations. The data processing process is relatively complex, requiring a large amount of computing resources on the device to support model training and inference, making it difficult for most devices with weaker computing capabilities to implement this technology.

[0005] Summary of the Invention

[0006] The present application provides a data processing method, which deploys machine learning models with the same structure on different devices, and jointly completes data processing by multiple devices to continuously improve the quality of the obtained data, reduce the data processing pressure of each device, and ensure that devices with weaker computing power can also realize data processing.

[0007] The first aspect of the present application provides a data processing method. Taking the data processing method being executed by a first device as an example, the first device may be a terminal device or a network device, or a partial component in the terminal device or the network device (such as a processor, a chip or a chip system, etc.). Alternatively, the first device may also be a logic module and / or software that can realize all or part of the functions of the terminal device. Specifically, the method includes: the first device receives first data from the second device, and the first data is data processed by a first machine learning model; that is, the second device obtains the first data through the first machine learning model, and then sends the first data to the first device. Then, the first device processes the first data through the second machine learning model to obtain the second data. The structure of the first machine learning model is the same as that of the second machine learning model, and the first device and the second device are used to jointly execute data processing.

[0008] In some cases, the first device and the second device may also be referred to as distributed devices. Different distributed devices implement joint data processing by exchanging data.

[0009] Simply put, the second device processes the data through a machine learning model and sends the processed data to the first device; the first device then continues to process the data based on the machine learning model of the same structure, thereby enabling the two devices to jointly execute data processing.

[0010] In this solution, by deploying machine learning models on different devices, multiple devices jointly complete data processing to continuously improve the quality of the obtained data, reduce the data processing pressure of each device, and ensure that devices with weaker computing power can also obtain data of the required quality.

[0011] In one possible implementation, the second machine learning model is a diffusion model. The data processing process using the diffusion model can be modeled as a Markov chain, and the second machine learning model is used to perform noise reduction on the first data. The diffusion model can be implemented using a neural network, such as a fully connected neural network, a convolutional neural network, or a residual neural network. The process of processing data using a diffusion model involves continuously processing the data output from the previous data processing by the diffusion model using the diffusion model, thereby achieving multi-step data processing based on the same diffusion model, ultimately obtaining high-quality output data.

[0012] In this solution, based on the characteristics of the diffusion model, the diffusion model of the same structure is deployed on different devices, and the different devices implement joint data processing in series. On the basis of obtaining high-quality data, the data processing pressure of each device is reduced.

[0013] In one possible implementation, the method further includes: the first device receiving first information from the second device, the first information being used to request the first device to process the first data. For example, when the second device needs to obtain high-quality data and is unable to complete the entire data processing process on its own, the second device partially processes the raw data using the first machine learning model to obtain the first data; and the second device sends the first data and the first information to the first device, requesting the first device to assist the second device in continuing to process the data.

[0014] In this solution, by interactively requesting information between devices, devices with weaker data processing capabilities can request assistance from other devices to complete data processing, fully utilizing the data processing capabilities of each device and ensuring that devices with weaker data processing capabilities can also obtain data of the required quality.

[0015] In one possible implementation, the first information sent by the second device to the first device indicates that the number of times the first data is to be processed is a first number. The first device processes the first number of times using the second machine learning model to obtain the second data. The capabilities of the first device support completing the first number of times of processing the first data.

[0016] In one possible implementation, since the second device requests assistance from the first device to complete data processing, after the first device processes the second data, the first device sends the second data to the second device to feedback the processed second data to the second device.

[0017] Alternatively, if the second device is not the first device to request assistance in data processing, the first information also indicates information about the source device. The source device is the first device to request assistance in data processing, the second device is one of the devices assisting in data processing, and the first device is another device assisting in data processing. In this way, after the first device processes the second data, it sends the second data to the source device, ensuring that the source device receives the final processed data.

[0018] In one possible implementation, the second device sends first information to the first device indicating that the number of times the first data is to be processed is a first number. The first device processes the first data using a second machine learning model for the second number to obtain second data, where the first number is greater than the second number, and the capabilities of the first device do not support completing the first number of processing of the first data. Furthermore, the first device sends the second data and the second information to a third device; the second information indicates that the number of times the second data is to be processed is a third number, where the third number is the difference between the first number and the second number, and the third device is configured to assist the first device in processing the data.

[0019] Simply put, suppose the second device requests the first device to assist in processing the first data 1,000 times, but the capability of the first device can only support the first device to complete processing of the first data 600 times. The first device then completes processing of the first data 600 times to obtain the second data; and the first device sends a second message to the third device to request the third device to continue processing the second data 400 times.

[0020] In this solution, multiple devices jointly process data, each completing a portion of the processing based on its own capabilities. Any unfinished data is then sent to the next device, which then continues processing. This approach leverages the data processing capabilities of each device while also coordinating the data processing efforts of multiple devices. This fully utilizes the processing capabilities of each device and ensures that even devices with less computing power can obtain the required data quality.

[0021] In a possible implementation, the method further includes: the first device sending assistance request information to the second device, where the assistance request information is used to request the second device to assist in processing data.

[0022] That is, the first device can actively request the second device to assist in processing data, and the second device processes the data first and then passes the processed data to the first device for further processing, avoiding the exchange of data twice between the first device and the second device.

[0023] In one possible implementation, the method further includes: the first device sending third information to the central device, the third information indicating the number of data processing times required by the first device; and the first device receiving feedback from the central device, the feedback information indicating that the second device is an assisting node. In other words, the first device may first provide feedback to the central device indicating the number of data processing times required by the first device, and the central device then coordinates the data among multiple devices, i.e., the central device instructs the second device to assist the first device in data processing.

[0024] Alternatively, the second device may provide the central device with the number of times it needs to process data. In this case, the central device can determine that the second device processes the data first, and then the first device continues processing the data based on the data processed by the second device. This effectively utilizes the data processed by the second device and improves the data processing efficiency of the first device.

[0025] In this solution, the central device is used to coordinate the joint data processing among the distributed devices, and the data processing tasks of each distributed device can be determined based on the needs of each distributed device, thereby improving the efficiency of the joint data processing of the distributed devices.

[0026] In one possible implementation, the method further includes: the first device receiving fourth information from the central device, the fourth information being used to indicate a number of times the first device needs to process data received from the second device. The first device processes the first data using a second machine learning model based on the fourth information to obtain second data required by the first device.

[0027] That is to say, in the process of the central device coordinating the distributed devices to jointly process data, the central device can determine the data processing order and the number of data processing times between each distributed device based on the data processing requirements of each distributed device, so that each distributed device can determine the number of data processing times after receiving the data sent by other distributed devices.

[0028] In one possible implementation, the fourth information also indicates information about a third device, which is the device to receive the data processed by the first device. The first device sends the second data to the third device based on the fourth information. That is, after the first device processes the second data with the assistance of the second device, it can send the second data to the third device, allowing the third device to use the second data or continue processing the second data.

[0029] Simply put, the central device can indicate in the information fed back to the distributed devices which distributed device the distributed device needs to receive data from, the number of times the data needs to be processed after receiving it, and which distributed device the processed data needs to be sent to, thereby effectively coordinating the joint data processing between the distributed devices.

[0030] In one possible implementation, the method further includes: the first device receiving fifth information from the second device, the fifth information indicating a number of times the first data has been processed. The first device processes the first data using a second machine learning model based on the number of times the first data has been processed and a number of times the data needs to be processed by the first device, to obtain the second data needed by the first device.

[0031] In this solution, when the previous distributed device sends processed data to the next distributed device, the previous distributed device indicates the number of times it has processed the data, which makes it easier for the next distributed device to determine the number of times the data still needs to be processed based on the number of times the data has been processed, thereby ensuring the joint processing of the data. There is no need for the central device to specify the number of data processing times, which is beneficial for each distributed device to dynamically adjust the number of data processing times according to actual operating conditions.

[0032] A second aspect of the present application provides a data processing method. Taking the data processing method as an example, the first device may be a terminal device or a network device, or a component (e.g., a processor, a chip, or a chip system) in the terminal device or network device. Alternatively, the first device may be a logic module and / or software capable of implementing all or part of the functions of the terminal device.

[0033] Specifically, the method includes: first, a first device processes raw data using a first machine learning model to obtain first data. Then, the first device sends the first data to a second device. Finally, the first device receives second data sent by the second device or another device, where the second data is obtained by processing based on a second machine learning model, and the structure of the first machine learning model is the same as that of the second machine learning model.

[0034] Optionally, the first device and the second device may interact in advance so that after receiving the data sent by the first device, the second device can determine that the data sent by the first device needs to be processed a certain number of times.

[0035] In other words, the first device first determines the number of times the original data needs to be processed, and then processes the original data a certain number of times to obtain the first data. Because the number of times the first device processes the original data is less than the required number of times, the first device sends the first data to the second device, implicitly requesting the second device to assist in processing the first data.

[0036] In this solution, by deploying machine learning models on different devices, data processing is completed jointly by the equipment of multiple devices to continuously improve the quality of the obtained data, reduce the data processing pressure of each device, and ensure that devices with weaker computing power can also obtain data of the required quality.

[0037] In one possible implementation, the first machine learning model is a diffusion model, the data processing process using the diffusion model can be modeled as a Markov chain, and the first machine learning model is used to perform noise reduction on the original data.

[0038] In one possible implementation, the first device may also send first information to the second device, where the first information is used to request the second device to perform processing on the first data, and / or the first information is used to indicate the number of times the first data is to be processed, where the number of times the first data is to be processed is determined based on the number of times the original data needs to be processed and the number of times the first device performs processing on the original data.

[0039] That is, in addition to sending the first data to be processed to the second device, the first device also sends the first information to the second device to instruct the second device how to process the first data.

[0040] A third aspect of the present application provides a data processing method. For example, the data processing method is performed by a central device. The central device can be a terminal device or a network device, or a component (e.g., a processor, chip, or chip system) in the terminal device or network device. Alternatively, the central device can be a logic module and / or software that can implement all or part of the terminal device's functions.

[0041] Specifically, the method includes: a central device receiving first information from a first device and second information from a second device, wherein the first information is used to indicate a first number of processing times of data required by the first device, and the second information is used to indicate a second number of processing times of data required by the second device, and the data processing model corresponding to the first number of processing times is the same as the data processing model corresponding to the second number of processing times; and the central device sending third information to the second device, the third information being used to instruct the second device to send processed data to the first device, wherein the second number of processing times of data required by the second device is less than or equal to the first number of processing times of data required by the first device.

[0042] Simply put, the first and second devices can also be called distributed devices. Each of them reports the number of times they have processed the data they requested to the central device. Based on the number of times they have processed the data, the central device determines the order in which they should be processed during the joint data processing process. This instructs the second device to send the processed data to the first device. In other words, the second device processes the data first and then sends it to the first device.

[0043] In this solution, the central device is used to coordinate the joint data processing among the distributed devices, and the data processing tasks of each distributed device can be determined based on the needs of each distributed device, thereby improving the efficiency of the joint data processing of the distributed devices.

[0044] In a possible implementation, the method further includes: the central device sending fourth information to the first device, where the fourth information is used to indicate the number of times the first device needs to process the data received from the second device.

[0045] For example, assuming that the first device requires data to be processed 1000 times and the second device requires data to be processed 600 times, the fourth information sent by the central device to the first device may indicate that the first device needs to process the data received from the second device 400 times.

[0046] A fourth aspect of the present application provides a model training method, which is applied to a first device in a training system. The first device can be a terminal device or a network device, or a partial component in the terminal device or network device (such as a processor, a chip, or a chip system, etc.). Alternatively, the first device can also be a logic module and / or software that can implement all or part of the terminal device functions. The training system includes multiple distributed devices. Specifically, the method includes: first, the first device obtains a training sample set, the training sample set including first data and second data, the first data is obtained based on the second data, and the second data is a training label for the first data. Then, the first device trains a first machine learning model based on the training sample set to obtain a trained first machine learning model. During the training process, the first data is used as input data for the first machine learning model, and the first machine learning model processes the first data, and calculates a loss function based on the processed data and the second data, and updates the parameters of the first machine learning model based on the loss function, thereby achieving training of the first machine learning model. After the training is completed, the first device sends the trained first machine learning model to the second device, which is a device for aggregating machine learning models with the same structure and different parameters trained by multiple devices. The second device can also be referred to as an aggregation device.

[0047] In this solution, by deploying the same machine learning model on different distributed devices and training the same machine learning model, the aggregation device finally aggregates the machine learning models trained by each distributed device, so that the training process can be split into different distributed devices for execution, thereby reducing the model training pressure of each distributed device.

[0048] In one possible implementation, the method further includes: the first device sends first information to the third device, and the first information is used to indicate the capabilities related to model training on the first device. The third device is used to determine the training content that multiple devices are responsible for based on the capabilities of multiple devices participating in the machine learning model training, so the third device can also be called a central device. After sending the first information, the first device receives second information from the third device, and the second information is used to indicate the number of times the first machine learning model trained on the first device processes the input data. The second information is also used to indicate the input data requirements of the first machine learning model. For example, in a case where the input data of the first machine learning model is obtained by performing noise processing on the target data, the input data requirements of the first machine learning model can be the number of times the target data is subjected to noise processing.

[0049] The process of the first device acquiring the training sample set includes: the first device processing the target data according to the input data requirement indicated by the second information and the number of times the first machine learning model processes the input data, thereby obtaining the second data and the first data. For example, assuming that the input data requirement is to perform MN to M noise addition processes on the target data to obtain a set of data, and then obtain the first data and the second data from the set, wherein the second data is the training label of the first data, and the number of noise addition processes required to obtain the second data is less than the number of noise addition processes required to obtain the corresponding first data.

[0050] In one possible implementation, the method also includes: the first device receives a second machine learning model from the second device; the first device trains the second machine learning model based on a training sample set to obtain a trained second machine learning model; and the first device sends the trained second machine learning model to the second device.

[0051] That is, after the second device aggregates the machine learning models of multiple distributed devices, the second device continues to send the aggregated second machine learning model to the first device, so that the first device continues to train the second machine learning model.

[0052] The fifth aspect of the present application provides a model training method, which is applied to a first device, which may be a terminal device or a network device, or a partial component in the terminal device or the network device (such as a processor, a chip or a chip system, etc.). Alternatively, the first device may also be a logic module and / or software that can realize all or part of the functions of the terminal device. The method includes: the first device receives a plurality of capability information, the plurality of capability information comes from a plurality of different devices, and each capability information in the plurality of capability information is used to indicate the capability related to model training on the device. Then, the first device sends different training configuration information to a plurality of different devices according to the plurality of capability information, the training configuration information is used to indicate the number of times the machine learning model trained on the device processes the input data, and the training configuration information is also used to indicate the input data requirements of the machine learning model trained on the device, and the machine learning models trained by the plurality of different devices are models with the same structure. Among them, the first device can also be called a central device.

[0053] Simply put, the process of continuously processing data using a machine learning model can be considered a Markov chain, and a single data processing operation using a machine learning model can be considered a link in the Markov chain. The central device can split the Markov chain into multiple sub-chains and allocate these sub-chains to different distributed devices based on their capabilities. This means that different distributed devices are used to perform different training tasks.

[0054] Optionally, the capabilities related to model training on the distributed device may include computing power, storage capacity, and communication capacity of the distributed device. Computing power can be measured by the number of operations per second that the distributed device can perform; storage capacity can be measured by the size of the storage space allocated to model training on the distributed device; and communication capacity can be measured by the data transmission rate allocated to the model training process on the distributed device. In addition to the various capabilities mentioned above, the capabilities related to model training on the distributed device may also include other capabilities that can affect model training, which are not specifically limited here.

[0055] In a sixth aspect of the present application, a communication device is provided, which includes: a transceiver module for receiving first data from a second device, where the first data is data processed by a first machine learning model; a processing module for processing the first data through a second machine learning model to obtain second data, where the structure of the first machine learning model is the same as that of the second machine learning model, and the communication device and the second device are used to jointly perform data processing.

[0056] In one possible implementation, the second machine learning model is a diffusion model, the data processing process using the diffusion model can be modeled as a Markov chain, and the second machine learning model is used to perform noise reduction on the first data.

[0057] In a possible implementation, the transceiver module is further configured to receive first information from the second device, where the first information is used to request the communication device to process the first data.

[0058] In one possible implementation, the first information is used to indicate that the number of times the first data is to be processed is the first number; the processing module is also used to process the first data the first number through a second machine learning model to obtain second data, wherein the capability of the first device supports the completion of the first number of processing of the first data.

[0059] In one possible implementation, the transceiver module is further used to send the second data to the second device; or, the transceiver module is used to send the second data to the source device, wherein the first information is also used to indicate information of the source device, and the source device is the first device to request assistance in processing data.

[0060] In one possible implementation, the first information is used to indicate that the number of times the first data is to be processed is the first number; the processing module is also used to process the first data for the second number through a second machine learning model to obtain second data, wherein the first number is greater than the second number, and the capability of the first device does not support the completion of the first number of processing of the first data; the transceiver module is also used to send the second data and the second information to a third device; wherein the second information is used to indicate that the number of times the second data is to be processed is the third number, and the third number is the difference between the first number and the second number, and the third device is used to assist the communication device in performing data processing.

[0061] In a possible implementation, the transceiver module is further configured to send assistance request information to the second device, where the assistance request information is used to request the second device to assist in processing data.

[0062] In one possible implementation, the transceiver module is further used to send third information to the central device, where the third information is used to indicate the number of data processing times required by the communication device; the transceiver module is further used to receive feedback information from the central device, where the feedback information is used to indicate that the second device is an assisting node.

[0063] In one possible implementation, the transceiver module is further used to receive fourth information from the central device, where the fourth information is used to indicate the number of times the communication device needs to process the data received from the second device; the processing module is further used to process the first data through a second machine learning model based on the fourth information to obtain the second data required by the communication device.

[0064] In a possible implementation, the fourth information is also used to indicate information of a third device, where the third device is a device to receive data processed by the first device; the transceiver module is further used to send the second data to the third device according to the fourth information.

[0065] In one possible implementation, the transceiver module is further used to receive fifth information from the second device, where the fifth information is used to indicate the number of times the first data has been processed; the processing module is further used to process the first data through a second machine learning model based on the number of processing times and the number of processing times of the data required by the communication device to obtain the second data required by the communication device.

[0066] In a possible implementation, the transceiver module is a transceiver, and the processing module is a processor.

[0067] In a seventh aspect, the present application provides a communication device, comprising: a processing module, used to process original data through a first machine learning model to obtain first data; a transceiver module, used to send the first data to a second device; the transceiver module is also used to receive second data sent by the second device or other devices, and the second data is obtained by processing the first data based on the second machine learning model, and the structure of the first machine learning model is the same as the structure of the second machine learning model.

[0068] In one possible implementation, the first machine learning model is a diffusion model, the data processing process using the diffusion model can be modeled as a Markov chain, and the first machine learning model is used to perform noise reduction on the original data.

[0069] In one possible implementation, the transceiver module is also used to send first information to the second device, where the first information is used to request the second device to perform processing on the first data, and / or the first information is used to indicate the number of times the first data is to be processed, and the number of times the first data is to be processed is determined based on the number of times the original data needs to be processed and the number of times the first device performs processing on the original data.

[0070] In a possible implementation, the transceiver module is a transceiver, and the processing module is a processor.

[0071] In an eighth aspect, the present application provides a communication device, comprising: a transceiver module for receiving first information from a first device and second information from a second device, the first information being used to indicate a first processing number of data required by the first device, the second information being used to indicate a second processing number of data required by the second device, and the data processing model corresponding to the first processing number is the same as the data processing model corresponding to the second processing number; a transceiver module for sending third information to the second device, the third information being used to instruct the second device to send processed data to the first device, wherein the second processing number of the data required by the second device is less than or equal to the first processing number of the data required by the first device.

[0072] In a possible implementation, the transceiver module is further configured to send fourth information to the first device, where the fourth information is configured to indicate a number of times the first device needs to process data received from the second device.

[0073] In a possible implementation, the transceiver module is a transceiver.

[0074] In a ninth aspect of the present application, a model training device is provided, comprising a transceiver module for acquiring a training sample set, the training sample set comprising first data and second data, the first data being obtained based on the second data, and the second data being a training label for the first data; a processing module for training a first machine learning model based on the training sample set to obtain a trained first machine learning model, wherein the first machine learning model is used to process the first data; a sending module for sending the trained first machine learning model to a second device, the second device being a device for aggregating machine learning models with the same structure and different parameters trained by multiple devices.

[0075] In one possible implementation, the sending module is further used to send first information to a third device, where the first information is used to indicate capabilities related to model training on the model training device, and the third device is used to determine the training content that multiple devices are responsible for based on the capabilities of multiple devices participating in the machine learning model training; the transceiver module is further used to receive second information from the third device, where the second information is used to indicate the number of times the first machine learning model trained on the model training device processes the input data, and the second information is also used to indicate the input data requirements of the first machine learning model; the processing module is further used to: process the original data according to the input data requirements indicated by the second information and the number of times the first machine learning model processes the input data to obtain second data; and process the second data according to the number of times the first machine learning model processes the input data indicated by the second information to obtain first data.

[0076] In one possible implementation, the transceiver module is also used to receive a second machine learning model from a second device; the processing module is also used to train the second machine learning model based on a training sample set to obtain a trained second machine learning model; and the sending module is also used to send the trained second machine learning model to the second device.

[0077] In a possible implementation, the transceiver module is a transceiver, and the processing module is a processor.

[0078] In the tenth aspect of the present application, a model training device is provided, including a transceiver module for receiving multiple capability information, the multiple capability information coming from multiple different devices, and each capability information in the multiple capability information is used to indicate the capability related to model training on the device; the transceiver module is used to send different training configuration information to multiple different devices according to the multiple capability information, the training configuration information is used to indicate the number of times the machine learning model trained on the device processes the input data, and the training configuration information is also used to indicate the input data requirements of the machine learning model trained on the device, and the machine learning models trained by multiple different devices are models with the same structure.

[0079] In a possible implementation, the transceiver module is a transceiver.

[0080] An eleventh aspect of an embodiment of the present application provides a communication device, comprising at least one processor, which is coupled to a memory; the memory is used to store programs or instructions; the at least one processor is used to execute the program or instructions, so that the device implements the method described in the first aspect or any possible implementation of the first aspect, or, so that the device implements the method described in the second aspect or any possible implementation of the second aspect, or, so that the device implements the method described in the third aspect or any possible implementation of the third aspect, or, so that the device implements the method described in the fourth aspect or any possible implementation of the fourth aspect, or, so that the device implements the method described in the fifth aspect or any possible implementation of the fifth aspect.

[0081] In a possible implementation, the communication device further includes the aforementioned memory. Optionally, the memory and the processor are integrated together, or the memory and the processor are separately provided.

[0082] In a possible implementation, the communication device further includes a transceiver for sending and receiving data or signaling.

[0083] A twelfth aspect of an embodiment of the present application provides a computer-readable storage medium storing one or more computer-executable instructions. When the computer-executable instructions are executed by a processor, the processor executes the method described in the first aspect or any possible implementation of the first aspect, or the processor executes the method described in the second aspect or any possible implementation of the second aspect, or the processor executes the method described in the third aspect or any possible implementation of the third aspect, or the processor executes the method described in the fourth aspect or any possible implementation of the fourth aspect, or the processor executes the method described in the fifth aspect or any possible implementation of the fifth aspect.

[0084] A thirteenth aspect of an embodiment of the present application provides a computer program product (or computer program) storing one or more computers. When the computer program product is executed by the processor, the processor executes the method as described in the first aspect or any possible implementation of the first aspect, or the processor executes the method as described in the second aspect or any possible implementation of the second aspect, or the processor executes the method as described in the third aspect or any possible implementation of the third aspect, or the processor executes the method as described in the fourth aspect or any possible implementation of the fourth aspect, or the processor executes the method as described in the fifth aspect or any possible implementation of the fifth aspect.

[0085] A fourteenth aspect of an embodiment of the present application provides a chip system, which includes at least one processor, for supporting a communication device to implement the functions involved in the above-mentioned first aspect or any possible implementation of the first aspect, or for supporting a communication device to implement the functions involved in the above-mentioned second aspect or any possible implementation of the second aspect, or for supporting a communication device to implement the functions involved in the above-mentioned third aspect or any possible implementation of the third aspect, or for supporting a communication device to implement the functions involved in the above-mentioned fourth aspect or any possible implementation of the fourth aspect, or for supporting the method described in the above-mentioned fifth aspect or any possible implementation of the fifth aspect of the communication device.

[0086] In one possible design, the chip system may further include a memory for storing program instructions and data necessary for the communication device. The chip system may be composed of a chip or may include a chip and other discrete components. Optionally, the chip system also includes an interface circuit that provides program instructions and / or data to the at least one processor.

[0087] A fifteenth aspect of an embodiment of the present application provides a communication system, which includes the communication device involved in the sixth and seventh aspects above, and / or, the communication system includes the communication device involved in the sixth, seventh and eighth aspects above, and / or, the communication system includes the communication device involved in the ninth and tenth aspects above.

[0088] Among them, the technical effects brought about by any design method in the sixth to fifteenth aspects can be referred to the technical effects brought about by the different implementation methods in the above-mentioned first to fifth aspects, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] FIG1 is a schematic diagram of a process of processing data using a diffusion model according to an embodiment of the present application;

[0090] FIG2 is a schematic diagram of a partial structure of a fully connected neural network provided in an embodiment of the present application;

[0091] FIG3 is a schematic diagram of a neural network training process provided by an embodiment of the present application;

[0092] FIG4 is a schematic diagram of a process of performing back propagation in a neural network according to an embodiment of the present application;

[0093] FIG5 is a schematic diagram of the architecture of a wireless communication system provided in an embodiment of the present application;

[0094] FIG6 is a schematic diagram of the architecture of a smart home communication system provided in an embodiment of the present application;

[0095] FIG7 is a flow chart of a model training method provided in an embodiment of the present application;

[0096] FIG8 is another flow chart of a model training method provided in an embodiment of the present application;

[0097] FIG9A is a flow chart of a data processing method 900 provided in an embodiment of the present application;

[0098] FIG9B is another flowchart of a data processing method 900 provided in an embodiment of the present application;

[0099] FIG10A is a flow chart of a data processing method 1000 provided in an embodiment of the present application;

[0100] FIG10B is another flowchart of a data processing method 1000 provided in an embodiment of the present application;

[0101] FIG11 is a flow chart of a data processing method 1100 provided in an embodiment of the present application;

[0102] FIG12 is a flow chart of a data processing method 1200 provided in an embodiment of the present application;

[0103] FIG13 is a schematic structural diagram of a communication device 1300 provided in an embodiment of the present application;

[0104] FIG14 is a schematic structural diagram of a model training device 1400 provided in an embodiment of the present application;

[0105] FIG15 is a schematic structural diagram of a communication device 1500 provided in an embodiment of the present application;

[0106] FIG16 is a schematic structural diagram of a communication device 1600 provided in an embodiment of the present application;

[0107] FIG17 is a schematic structural diagram of a communication device 1700 provided in an embodiment of the present application. DETAILED DESCRIPTION

[0108] The following describes the embodiments of the present application in conjunction with the accompanying drawings. Obviously, the embodiments described are only part of the embodiments of the present application, rather than all the embodiments. Those skilled in the art will appreciate that with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0109] The terms "first," "second," and the like in the specification and claims of this application and the accompanying drawings are used to distinguish similar objects (e.g., to distinguish objects in the same embodiment) and are not necessarily used to describe a specific order or precedence. In addition, the objects defined by "first," "second," and the like (e.g., "first information," "first device," "second information," "second device," and the like) may refer to different objects in different embodiments. For example, the "first device" in the first embodiment may refer to a distributed node, and the "first device" in the second embodiment may refer to a central node. It should be understood that the terms used in this manner are interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein.

[0110] In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or device that includes a series of steps or modules is not necessarily limited to those steps or modules explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to the process, method, product, or device. The naming or numbering of steps in this application does not mean that the steps in the method flow must be executed in the time / logical sequence indicated by the naming or numbering. The named or numbered process steps can be executed in a different order according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved.

[0111] To facilitate understanding, the technical terms involved in the embodiments of the present application are first introduced below.

[0112] (1) Terminal equipment

[0113] A terminal device is a wireless terminal device that can receive scheduling information and instruction information sent by a network device. A wireless terminal device can be a device that provides voice and / or data connectivity to a user, a handheld device with wireless connection capabilities, or other processing devices connected to a wireless modem.

[0114] The terminal device can communicate with one or more core networks or the Internet via a wireless access network (RAN). For example, the terminal device can be a mobile terminal device, such as a mobile phone (also known as a "cellular" phone, mobile phone), a computer, and a data card. For example, it can be a portable, pocket-sized, handheld, computer-built-in, or vehicle-mounted mobile device that exchanges voice and / or data with the wireless access network. For another example, the terminal device can also be a personal communication service (PCS) phone, a cordless phone, a Session Initiation Protocol (SIP) phone, a wireless local loop (WLL) station, a personal digital assistant (PDA), a tablet personal computer (Tablet PC), a computer with wireless transceiver capabilities, and other devices. Generally, the terminal device can also be called a system, a subscriber unit, a subscriber station, a mobile station, a mobile station (MS), a remote station, an access point (AP), a remote terminal, an access terminal, a user terminal, a user agent, a subscriber station (SS), a customer premises equipment (CPE), a terminal, a user equipment (UE), a mobile terminal (MT), etc.

[0115] As an example and not a limitation, in the embodiments of the present application, the terminal device may also be a wearable device. Wearable devices may also be referred to as wearable smart devices or smart wearable devices, etc., which are a general term for wearable devices that are intelligently designed and developed using wearable technology for daily wear, such as glasses, gloves, watches, clothing, and shoes. A wearable device is a portable device that is worn directly on the body or integrated into the user's clothes or accessories. Wearable devices are not only hardware devices, but also achieve powerful functions through software support, data interaction, and cloud interaction. Broadly speaking, wearable smart devices include those that are fully functional, large in size, and can achieve complete or partial functions without relying on smartphones, such as smart watches or smart glasses, etc., as well as those that only focus on a certain type of application function and need to be used in conjunction with other devices such as smartphones, such as various smart bracelets, smart helmets, and smart jewelry for vital sign monitoring.

[0116] The terminal may also be a drone, a robot, a terminal in device-to-device (D2D) communication, a terminal in vehicle-to-everything (V2X), a virtual reality (VR) terminal device, an augmented reality (AR) terminal device, a wireless terminal in industrial control, a wireless terminal in self-driving, a wireless terminal in remote medical, a wireless terminal in smart grid, a wireless terminal in transportation safety, a wireless terminal in smart city, a wireless terminal in smart home, etc.

[0117] In addition, the terminal device may also be a terminal device in a fifth-generation (5G) communication system, a later-evolved communication system (e.g., a sixth-generation (6G) communication system), or a terminal device in a future-evolved public land mobile network (PLMN). For example, a 6G network may further extend the form and functionality of a 5G communication terminal, and 6G terminals include but are not limited to vehicles, cellular network terminals (with integrated satellite terminal functions), drones, and Internet of Things (IoT) devices.

[0118] In an embodiment of the present application, the above-mentioned terminal device may have AI processing capabilities and can use AI models to process data.

[0119] (2) Network equipment

[0120] A network device may refer to a device that provides wireless access services in a wireless network. For example, a network device may be a RAN node (or RAN device) that connects a terminal device to a wireless network, and may also be referred to as a base station. Currently, some examples of RAN devices include: the next generation Node B (gNB) in a 5G communication system, a transmission reception point (TRP), an evolved Node B (eNB), a radio network controller (RNC), a Node B (NB), a home base station (e.g., home evolved Node B, or home Node B, HNB), a baseband unit (BBU), or a wireless fidelity (Wi-Fi) access point (AP). In addition, in a network structure, a network device may include a centralized unit (CU) node, a distributed unit (DU) node, or a RAN device including a CU node and a DU node.

[0121] The network device may be any other device that provides wireless communication functionality to the terminal device. The embodiments of this application do not limit the specific technology and device form used by the network device. For ease of description, the embodiments of this application do not limit this.

[0122] The network equipment may also include core network equipment, which may include, for example, a mobility management entity (MME), a home subscriber server (HSS), a serving gateway (S-GW), a policy and charging rules function (PCRF), and a public data network gateway (PDN gateway, P-GW) in a fourth generation (4G) network; and network elements such as an access and mobility management function (AMF), a user plane function (UPF), or a session management function (SMF) in a 5G network. In addition, the core network equipment may also include other core network equipment in a 5G network and the next generation or future network of the 5G network.

[0123] In an embodiment of the present application, the above-mentioned network device may also have a network node with AI capabilities, which can provide AI services for terminals or other network devices. For example, it can be an AI node on the network side (access network or core network), a computing power node, a RAN node with AI capabilities, a core network element with AI capabilities, etc.

[0124] In an embodiment of the present application, the device for implementing the function of the network device may be a network device, or a device that can support the network device to implement the function, such as a chip system, which may be installed in the network device.

[0125] (3) Markov Chain (MC)

[0126] A Markov chain is a stochastic process in probability theory and mathematical statistics that has the Markov property and exists in a discrete index set and state space.

[0127] Simply put, a Markov chain is a set of discrete random variables with the Markov property. Specifically, for a set X of random variables in the probability space (Ω, F, P) with a one-dimensional countable set as the exponent set, where the probability space (Ω, F, P) is a measure space with a total measure of 1; Ω is a non-empty set, also called a sample space; F is a non-empty subset of the power set of the sample space Ω; and P is the probability. If the values ​​of the random variables in the random variable set X are all in the countable set: X i =s i , s i ∈s, and the conditional probability of the random variable satisfies the relationship shown in Formula 1:

[0128] p(X t+1 |X t ,...,X1)=p(X t+1 |X t ) Formula 1

[0129] Then, X is called a Markov chain, the countable set S is called the state space, and the value of the Markov chain in the state space is called the state. i is a random variable in the random variable set X, i can be any integer greater than 0; S i is an element in a countable set S. A countable set S is a set in which each element can establish a one-to-one correspondence with each element of the natural number set.

[0130] (4) Diffusion model

[0131] The diffusion model is an artificial intelligence model that uses the principles of Markov chains to process data. For noisy data, the diffusion model can be used to reduce the noise, thereby obtaining higher-quality data. The following describes how the diffusion model reduces noise.

[0132] Please refer to Figure 1, which is a schematic diagram of the process of processing data using a diffusion model provided by an embodiment of the present application. As shown in Figure 1, given data X0 obeys the distribution q(X0), the data X0 is gradually denoised in the order from right to left in Figure 1. Each time the denoising is performed, a new data is obtained. After the data X0 is denoised T times, the data X1 to X T There are T data in total.

[0133] Specifically, the data noise process can be regarded as a Markov process. Based on the data X0, we can obtain X1 to X T The conditional probability is: Among them, the value range of t is 1≤t≤T, ∏ represents the product, and the single-step transition probability is Indicates that from X t-1 to X t It is a is the mean and β t is the Gaussian distribution transformation of the variance; the variance parameter β t is a designable parameter. In the process of data noise addition, when the number of noise addition steps T is large enough and β t When the choice is reasonable, the final X T will conform to a Gaussian distribution with mean 0 and variance 1, that is

[0134] In this case, the data can be denoised by a reverse process (i.e., the process of processing data from left to right in Figure 1). Specifically, starting from a sample obtained by sampling from a Gaussian distribution, data samples that conform to the given distribution are gradually generated. Since the transition probability q(X t |X t-1 ) is difficult to calculate, so it can be approximated by a neural network. For example, with p θ (X t |X t-1 ) approximates the transition probability of the reverse process and makes p θ (X t |X t-1 ) conforms to the Gaussian distribution, that is in, Indicates that from X t to X t-1 is a μ θ (Xt ， t) is the mean and Σ θ (X t , t) is the Gaussian distribution transformation of the variance; the mean μ θ (X t , t) and variance Σ θ (X t , t) can be approximated by neural networks. Experiments have found that the variance term has little influence and can generally be fixed, such as where σ t is the parameter that is not learned, that is, σ t is a designable parameter. The mean μ is learned through the neural network θ (X t , t), that is, the input of the neural network is the data X in the tth step t and step index t, the trainable parameter is θ.

[0135] Experiments show that neural networks can also be used to approximate the noise term ε θ (X t , t), where the noise term ε θ (Xt , t) represents the t-1 to X t The noise parameters are then Get the generated data X0, where α t =1-β t , When t=1, z=0; when t>1

[0136] In simple terms, as shown in Figure 1, by gradually adding noise to the original data X0, we can eventually obtain data X with noise that satisfies the Gaussian distribution. T By adopting the diffusion model, the data X T The data X0 can be restored by performing a gradual noise reduction process. The diffusion model is a neural network model, such as a fully connected neural network, a convolutional neural network, a residual neural network, etc. This embodiment does not limit the specific model structure of the diffusion model.

[0137] For ease of understanding, the following will take a fully connected neural network as an example to introduce the neural network and its training method. Among them, a fully connected neural network is also called a multilayer perceptron (MLP). Please refer to Figure 2, which is a schematic diagram of a partial structure of a fully connected neural network provided in an embodiment of the present application. As shown in Figure 2, an MLP includes an input layer (left side), an output layer (right side), and multiple hidden layers (middle).

[0138] Optionally, the data corresponding to the input layer in FIG2 may be referred to as input data, which may include data required for training or data required for inference, such as the data X in FIG1 above. t The multiple hidden layers in Figure 2 are configured with corresponding model parameters for processing the input data based on these model parameters. The data corresponding to the output layer in Figure 2 can be referred to as output data, which is the data obtained after the multiple hidden layers process the input data, such as the data X0 in Figure 1 above.

[0139] In addition, each layer of the above MLP contains several nodes, called neurons, where neurons in two adjacent layers are connected to each other.

[0140] Alternatively, considering neurons in two adjacent layers, the output h of the neurons in the next layer is the weighted sum of the outputs x of all neurons in the previous layer connected to it and is processed by the activation function, which can be expressed as the following formula 2:

[0141] h=f(wx+b) Formula 2

[0142] Among them, w is the weight matrix, b is the bias vector, and f is the activation function.

[0143] Furthermore, the output of the neural network can be expressed as the following formula 3:

[0144] y=f n (w n f n-1 (...)+b n ) Formula 3

[0145] Here, n is the number of layers in the neural network, and its value is an integer greater than 1. 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, and the process of obtaining this mapping from random w and b using existing data is called neural network training.

[0146] Optionally, the specific method of training is to evaluate the output results of the neural network using a loss function. Please refer to Figure 3, which is a schematic diagram of the process of neural network training provided in an embodiment of the present application. As shown in Figure 3, the error can be backpropagated, and the neural network parameters (including w and b) can be iteratively optimized by the gradient descent method until the loss function reaches a minimum value, that is, the "optimal point" in Figure 3. It can be understood that the neural network parameters corresponding to the "optimal point" in Figure 3 can be used as the neural network parameters in the trained AI model information.

[0147] Alternatively, the gradient descent process can be expressed as the following formula 4:

[0148]

[0149] Among them, θ is the parameter to be optimized (including w and b), L is the loss function, and η is the learning rate, which controls the step size of gradient descent. represents the partial derivative, is the partial derivative of the loss function L, is the partial derivative of the parameter to be optimized.

[0150] Furthermore, the chain rule for partial derivatives can be used in the back propagation process. Please refer to Figure 4, which is a schematic diagram of the back propagation process of a neural network provided by an embodiment of the present application. As shown in Figure 4, the gradient of the parameters of the previous layer can be recursively calculated by the gradient of the parameters of the next layer, which can be expressed as the following formula 5:

[0151]

[0152] Among them, W ij is the weight of node j connecting to node i, S i is the weighted sum of the inputs to node i.

[0153] The above introduces the technical terms involved in the embodiments of the present application. The following introduces the application scenarios of the methods provided in the embodiments of the present application.

[0154] Please refer to Figure 5, which is a schematic diagram of the architecture of a wireless communication system provided in an embodiment of the present application. As shown in Figure 5, the wireless communication system includes a network device 501 and a terminal device 502. Among them, the terminal device 502 may include one or more terminal devices, such as the smart bracelet, smart phone, smart TV and laptop shown in Figure 5. Among them, the network device 501 establishes a wireless connection with each terminal device in the terminal device 502, and a wireless connection can also be established between the terminal devices in the terminal device 502. In the wireless communication system shown in Figure 5, the network device 501 can send downlink data to the terminal device 502, such as a model that needs to be trained; each terminal device in the terminal device 502 can send uplink data to the network device 501, such as a trained model. In addition, the terminal devices in the terminal device 502 can also send data to each other, such as data required in the model training process or data required in the model inference process.

[0155] It should be noted that the wireless communication systems mentioned in the embodiments of the present application include but are not limited to: fifth generation mobile communication technology (5G) communication system, 6G communication system, satellite communication system, short-range communication system, narrowband Internet of Things system (NB-IoT), global system for mobile communications (GSM), enhanced data rate for GSM evolution system (EDGE), wideband code division multiple access system (WCDMA), code division multiple access 2000 system (CDMA2000), time division synchronous code division multiple access system (TD-SCDMA) and long term evolution system (LTE). The embodiments of the present application do not limit the specific architecture of the wireless communication system.

[0156] Please refer to Figure 6, which is a schematic diagram of the architecture of a smart home communication system provided in an embodiment of the present application. As shown in Figure 6, in a smart home scenario, various smart home products are connected through a wireless network to enable data transmission between smart home products. In Figure 6, taking smart home products such as smart TVs, smart air purifiers, smart water dispensers, smart speakers, and sweeping robots as examples, these smart home products are all connected to the same wireless network through a wireless router, thereby enabling data interaction between various smart home products. In addition to the smart home products in the above examples, other types of smart home products may also be included in actual applications, such as smart refrigerators, smart range hoods, smart curtains, and other smart home products. This embodiment does not limit the types of smart home products.

[0157] In addition, different smart home products can also be directly connected wirelessly without the need to access the same wireless network through a wireless router. For example, the smart home products can be connected wirelessly via Bluetooth.

[0158] In addition to the scenarios described in Figures 5 and 6 above, the method provided in the embodiments of the present application can also be applied to other communication system scenarios. For example, in a smart factory scenario, different devices (such as intelligent robots, lathes, transport vehicles, etc.) are connected via a wireless network and transmit data to each other via the wireless network. The embodiments of the present application do not limit the specific scenarios in which the data processing method is applied.

[0159] The above describes the application scenarios of the methods provided in the embodiments of the present application. For ease of understanding, the following will first introduce the model training method executed in the model training phase in the embodiments of the present application, and then further introduce the data processing method executed in the model inference phase in the embodiments of the present application.

[0160] Please refer to Figure 7, which is a flow chart of a model training method provided in an embodiment of the present application. As shown in Figure 7, the model training method includes the following steps 701-708.

[0161] Step 701: The central device receives multiple capability information.

[0162] In this embodiment, the central device is used to obtain the capability information of each distributed device, and assign training tasks to each distributed device based on the capability information of each distributed device. Among them, the multiple capability information received by the central device comes from multiple different distributed devices (for example, distributed device 1-distributed device N shown in Figure 7). Moreover, each capability information in the multiple capability information is used to indicate the capability related to model training on the distributed device. Simply put, each distributed device collects the capabilities related to model training on this device, and feeds back its own capabilities related to model training to the central device by sending capability information to the central device.

[0163] Specifically, the central device is, for example, the terminal device or network device described above; or, the central device is a device used to implement the functions of the terminal device or network device described above, such as a chip or chip system in the terminal device or network device. The distributed device is, for example, the terminal device described above, or a device used to implement the functions of the terminal device described above. In a specific example, the central device can be the base station described above, and the distributed device can be a terminal device such as a smart bracelet, smart watch, smart TV, smartphone, or laptop computer.

[0164] Optionally, the capabilities related to model training on the distributed device may include one or more capabilities of the distributed device, such as computing power, storage capacity, and communication capacity. Among them, computing power can be measured by the number of operations that the distributed device can perform per second; storage capacity can be measured by the size of the storage space allocated to model training on the distributed device; and communication capacity can be measured by the data transmission rate allocated to the model training process on the distributed device. In addition to the various capabilities mentioned above, the capabilities related to model training on the distributed device may also include other capabilities that can affect model training, which is not specifically limited in this embodiment.

[0165] In step 702, the central device sends corresponding training configuration information to a plurality of different distributed devices according to the plurality of capability information.

[0166] After receiving multiple capability information, the central device can obtain the model training capabilities of each distributed device participating in model training. Based on the model training capabilities of each distributed device, the central device can determine the training configuration information of each distributed device throughout the model training phase. The training configuration information refers to the specific training tasks that the distributed device needs to perform during the model training phase.

[0167] For example, as shown above in the introduction to the diffusion model, the process of using the diffusion model to reduce noise in data actually involves repeatedly processing the data using the same machine learning model, gradually reducing the noise in the data and ultimately producing high-quality data. The more times the diffusion model performs noise reduction on the data, the higher the quality of the resulting data.

[0168] Simply put, the process of continuously processing data using a machine learning model can be considered a Markov chain, and each data processing operation using a machine learning model can be considered a link in the Markov chain. The central device can split the Markov chain into multiple sub-chains and, based on the capabilities of each distributed device, allocate the resulting sub-chains to different distributed devices. After the central device allocates the resulting sub-chains to different distributed devices, the different distributed devices are used to perform different training tasks. For example, distributed device 1 performs training task 1, distributed device 2 performs training task 2, and distributed device 3 performs training task 3. Alternatively, some distributed devices can be used to perform the same training task. For example, distributed device 1 performs training task 1, distributed device 2 also performs training task 1, and distributed device 3 performs training task 2.

[0169] For example, the training configuration information sent by the central device indicates the number of times the machine learning model trained on the distributed devices processes input data. Furthermore, because each distributed device is responsible for different steps in the training phase, the quality requirements for the input data of the machine learning models trained on each distributed device also vary. Therefore, the training configuration information also indicates the input data requirements of the machine learning models trained on the distributed devices.

[0170] For example, suppose the training task in the entire training phase is: to use the machine learning model to process the data X that has been subjected to T noise processing. T (Input data) is processed T times to obtain data X0; then the training task can be divided into three different sub-training tasks, and these three sub-training tasks are deployed in different distributed devices. Among them, the first sub-training task can be the data X that has been processed T times with noise by the machine learning model. T Perform n processing to obtain data X T-n The second sub-training task can be to use the machine learning model to process the data X that has been subjected to Tm noise processing. T-m Perform k processing to obtain data X T-m-k ; The third sub-training task can be to use the machine learning model to process the data X with noise added w times w Perform w processing times to obtain data X0.

[0171] Step 703: The aggregation device sends the machine learning model and target data samples to multiple distributed devices.

[0172] In this embodiment, the aggregation device is used to aggregate the models trained by each distributed device and feed the aggregated model back to each distributed device. The aggregation device and the aforementioned central device can be the same device or different devices, which is not specifically limited in this embodiment.

[0173] Exemplarily, the aggregation device may be the terminal device or network device described above; alternatively, the aggregation device may be a device used to implement the functions of the terminal device or network device, such as a chip or chip system in the terminal device or network device. In a specific example, the central device may be the base station described above, the aggregation device may be the base station or server described above, and the distributed device may be a terminal device such as a smart bracelet, smart watch, smart TV, smartphone, or laptop computer.

[0174] Optionally, the machine learning model sent by the aggregator to multiple distributed devices is identical, allowing the multiple distributed devices to perform model training on the same machine learning model. Furthermore, the parameters in the machine learning model sent by the aggregator may be initial parameters obtained after random initialization. For example, the machine learning model sent by the aggregator may be a diffusion model used to perform noise reduction on data.

[0175] In addition, the target data samples sent by the aggregation device to multiple distributed devices may also be the same, so that different distributed devices can generate corresponding training sample sets according to the training configuration information. The target data samples sent by the aggregation device to multiple distributed devices may be data with relatively high quality, for example, the target data sample (such as X0 in FIG1 ) is an image without noise. The process of the distributed device training the machine learning model is as follows: the distributed device first performs noise processing on the target data sample to obtain a training sample of lower quality; then, the distributed device inputs the training sample into the machine learning model, performs noise reduction processing on the training sample through the machine learning model, and trains the machine learning model based on the output result of the machine learning model.

[0176] Optionally, the target data sample sent by the aggregation device can be determined by the type of machine learning model that needs to be trained. This embodiment does not specifically limit the specific type of the target data sample.

[0177] For example, when the machine learning model that needs to be trained is the machine learning model of each transceiver module in the communication system, such as a transmitter machine learning model, a receiver machine learning model, a channel estimation machine learning model, a channel compression feedback machine learning model, a precoding machine learning model, a beam management machine learning model or a positioning machine learning model, the data sample can be channel data.

[0178] For another example, when the machine learning model to be trained is an image processing model, such as an image classification model, an image enhancement model, an image compression model, or an image detection model, the data sample may be image data.

[0179] For another example, when the machine learning model to be trained is a speech processing model, such as a speech recognition model or a speech generation model, the data sample can be speech data.

[0180] It should be noted that, in this embodiment, the execution order of steps 702 and 703 is not specifically limited. Step 703 may be executed before step 702, or step 703 and step 702 may be executed simultaneously.

[0181] Step 704: The distributed device generates a training sample set based on the target data sample.

[0182] Since the training configuration information sent by the central device to the distributed device indicates the machine learning model trained on the distributed device for the input data (such as X in FIG1 T ) and the input data requirements, the distributed device can generate a training sample set based on the target data sample (e.g., X0 in FIG1 ) for subsequent training of the machine learning model. Each training sample in the training sample set generated by the distributed device satisfies the input data requirements indicated in the training configuration information.

[0183] Optionally, the training sample set generated by the distributed device based on the target data sample may include first data and second data, wherein the first data is the input data of the machine learning model during the training process, and the second data is the training label of the first data. The training label of the input data of the machine learning model is used to generate a loss function in combination with the output result of the machine learning model, so as to update the machine learning model based on the loss function and complete the training of the machine learning model. Taking the first data and the second data as an example, in a round of iterative training of the machine learning model, the first data is input into the machine learning model to obtain the output result of the machine learning model; then, the loss function is constructed by calculating the difference between the output result of the machine learning model and the second data (i.e., the training label of the input data); finally, the parameters in the machine learning model are updated based on the value of the loss function, thereby completing a round of iterative training of the machine learning model.

[0184] During the generation of the first and second data, the distributed device processes the target data sample (e.g., adds noise) based on the input data requirements indicated by the training configuration information and the number of times the machine learning model processes the input data, thereby obtaining the second data. The distributed device then processes the second data (e.g., adds noise) based on the number of times the machine learning model processes the input data indicated by the training configuration information, thereby obtaining the first data.

[0185] The input data requirement indicates the type of input data. For example, the input data is the data obtained by performing noise addition processing on the target data sample a specified number of times. After the distributed device processes the target data sample, the second data is obtained. The number of times the distributed device processes the target data sample is the difference between the number of times indicated in the input data requirement and the number of times the machine learning model processes the input data.

[0186] Then, the distributed device processes the second data according to the number of times the machine learning model indicated by the training configuration information processes the input data to obtain the first data.

[0187] For example, assuming that the input data requirement indicates that the input data is data obtained by performing noise addition processing on the target data sample M times, and the number of times the machine learning model processes the input data is N times, then the distributed device can perform noise addition processing on the target data sample X0 from MN times to M times to obtain data {X M-N , X M-N+1 ,…,X M}, the data X M-N As the second data, X M As the first data. That is, the first data (input data) is the data X after performing M times of noise addition processing on the target data sample X0. M The second data is the data X after performing MN times of noise addition processing on the target data sample X0. M-N , that is, the training label, is used in a round of iterative training of the machine learning model (such as noise reduction training) as an object for difference comparison with the output result obtained by inputting the first data into the machine learning model, so as to determine the loss function to update the parameters in the machine learning model, thereby completing a round of iterative training of the machine learning model.

[0188] That is to say, in actual applications, the distributed device first performs noise processing on the target data sample MN times based on the input data requirements indicated by the training configuration information and the number of times the machine learning model processes the input data to obtain the second data, and then performs noise processing on the second data N times to obtain the first data.

[0189] For example, assuming that the target data sample obtained by the distributed device from the aggregation device is data X0, the training configuration information indicates that the number of times the machine learning model trained on the distributed device processes the input data is 5, and the input data requirement of the machine learning model indicates that the input data is data after the target data sample is denoised 15 times. Then, the distributed device can first perform 10 to 15 denoising processes on the target data sample X0 obtained from the aggregation device to obtain data {X 10 , X 11 , X 12 , X 13 , X 14 , X 15}. Among them, data X 10 Refers to the data obtained by performing 10 noise addition processes on the target data sample X0 (i.e., the second data); data X 15 Refers to the data obtained by performing 15 noise addition processes on the target data sample X0 (i.e., the first data), data X 15 Can be X 10 The data is obtained after 5 times of noise processing. Then, during the training process, the data X 15 As the input data of the machine learning model (i.e. the first data mentioned above), and the data X 10 As the training label of the input data (i.e. the second data mentioned above). Optionally, the distributed device can perform noise processing on the data sample according to the formula The conditional probability distribution shown is used to sample and obtain the input data.

[0190] In step 705, the distributed device trains the machine learning model based on the training sample set to obtain a trained machine learning model.

[0191] For example, after generating a training sample set, the distributed device inputs the first data in the training sample set as input data into the machine learning model to obtain output data of the machine learning model. Then, the distributed device calculates a loss function based on the training label of the input data (i.e., the second data) and the output data, and updates the machine learning model based on the value of the loss function to obtain a trained machine learning model. The specific process of updating the machine learning model based on the loss function can be referred to the above description and will not be repeated here.

[0192] Step 706: The distributed device sends the trained machine learning model to the aggregate device.

[0193] After each distributed device trains the machine learning model based on the training samples generated on the device and obtains the trained machine learning model, each distributed device sends the trained machine learning model to the aggregation device so that the aggregation device can aggregate the machine learning model.

[0194] In step 707, the aggregation device aggregates the trained machine learning models sent by each distributed device to obtain an aggregated model.

[0195] Optionally, since the machine learning models trained by each distributed device are models with the same structure, the machine learning models trained by each distributed device are models with the same structure but different parameters. After the aggregation device receives multiple trained machine learning models sent by the distributed devices, the aggregation device can perform a weighted summation of the parameters in the multiple trained machine learning models to obtain new parameters, where the new parameters are the parameters of the aggregation model. In other words, after the aggregation device aggregates the trained machine learning models sent by each distributed device, the structure of the obtained aggregation model does not change, but the parameters in the aggregation model do change, and the parameters in the aggregation model are obtained based on the trained machine learning models sent by each distributed device.

[0196] It should be noted that in addition to the above-mentioned method of weighted summation of parameters in multiple trained machine learning models to obtain an aggregated model, the aggregation device can also achieve model aggregation through other methods, which is not specifically limited in this embodiment.

[0197] Step 708: The aggregation device sends the aggregation model to each distributed device.

[0198] After achieving model aggregation, the aggregation device sends the aggregated model to each distributed device so that each distributed device can continue to train the aggregated model.

[0199] It is understandable that the above steps 701-708 introduce the process of the distributed device performing the first round of iterative training on the machine learning model. In actual applications, the distributed device may need to perform multiple rounds of iterative training on the machine learning model. Therefore, after each distributed device receives the aggregated model sent by the aggregation device, each distributed device and the aggregation device cyclically executes the above steps 704-708 until the conditions for terminating the machine learning model training are met. Among them, the conditions for terminating the machine learning model training can be that the number of rounds of iterative training of the machine learning model by the distributed device reaches a preset number of rounds; or, the performance of the machine learning model trained by the distributed device meets the preset requirements.

[0200] For ease of understanding, the model training method provided by the embodiment of the present application will be described in detail below with reference to specific examples. As shown in Figure 8, Figure 8 takes a base station as an aggregation device and terminal devices such as smart phones, smart watches, smart bracelets, and laptops as multiple distributed devices as an example to describe the process of the model training method in detail. As shown in Figure 8, the process of the model training method includes the following four stages.

[0201] In stage 1, the base station sends target data samples X0 and machine learning models to multiple terminal devices participating in model training.

[0202] During the first round of iterative training of the machine learning model, the base station may send a machine learning model with randomly initialized parameters to multiple terminal devices. During the Nth round of iterative training (N is greater than 1), the base station may send a machine learning model to multiple terminal devices that is an aggregated model of multiple machine learning models obtained from the previous round of iterative training.

[0203] In stage 2, multiple terminal devices generate training samples based on the target data sample X0 and train the machine learning model based on the training samples.

[0204] Since terminal devices such as smartphones, smart watches, smart bracelets and laptops are assigned different training contents, each terminal device generates training samples that match the training content based on the target data samples, and trains the machine learning model sent by the base station based on the training samples.

[0205] For example, as shown in FIG8 , the input data requirement of the machine learning model trained in the smartphone indicates that the input data is the data obtained by performing T noise additions on the target data sample X0; and the machine learning model in the smartphone needs to process the input data 3 times to obtain the output data. Therefore, in the process of generating training samples, the smartphone performs T-3 to T noise additions on the target data sample X0 to obtain the data {X T-3 , X T-2 , X T-1 , X T}; Based on data {X T-3 , X T-2 , X T-1 , X T}, the smartphone can construct a set of training samples (X T , X T-3 ), where the data X in the training sample T is the input data, data X T-3The training labels are used. During training, the smartphone reuses the machine learning models to be trained, creating a master model consisting of three identical models connected in sequence. The input of each model in the master model is the output of the previous model. The smartphone then inputs the input data from the training samples into the master model to obtain the output data from the master model. This allows the smartphone to construct a loss function for updating the machine learning model based on the output data and the training labels from the training samples.

[0206] For example, for the training sample (X T , X T-3 ), the data X T-3 Input into the total model and get the output data X of the total model T ', according to the output data X T ' and training labels X T Calculate the value of the loss function. Then, update the parameters of the machine learning model based on the calculated loss function value.

[0207] Optionally, since the overall model is obtained by sequentially connecting three identical machine learning models, each machine learning model in the overall model actually has corresponding output data. When training a machine learning model, in addition to constructing a loss function based on the output data of the overall model (i.e., the output data of the third machine learning model in the overall model), a loss function can also be constructed based on the output data of other machine learning models in the overall model.

[0208] For example, the data X T-3 After inputting the total model, the output data X output by the first machine learning model in the total model is obtained T-2 ', the output data X output by the second machine learning model in the total model T-1 ', the output data X output by the third machine learning model in the total model T '. Then, based on the training label X T-2 And output data X T-2 'Construct loss function 1 based on training label X T-1 And output data X T-1 'Construct loss function 2 based on training label X T And output data X T 'Construct loss function 3. Finally, the total loss function is calculated based on loss function 1, loss function 2, and loss function 3, and the machine learning model is updated based on the value of the total loss function to obtain the trained machine learning model.

[0209] Similarly, for a smartwatch, the input data for the machine learning model trained in the smartwatch is X T-3, that is, the data obtained after the target data sample X0 is subjected to T-3 noise additions; and the machine learning model in the smartwatch needs to process the input data twice to obtain the output data. In the process of generating training samples, the smartwatch performs T-5 to T-3 noise additions on the target data sample X0 to obtain the data {X T-5 , X T-4 , X T-3}; Based on data {X T-5 , X T-4 , X T-3} obtain multiple sets of training samples, each set of training samples in the multiple sets of training samples includes input data and training labels, and the multiple sets of training samples are, for example, (X T-5 , X T-4 )、(X T-4 , X T-3 )、(X T-5 , X T-3 ). Referring to the process of training a machine learning model based on training data in a smartphone, a smartwatch trains a machine learning model based on the generated training data, and can also obtain a trained machine model.

[0210] For the smart bracelet, the input data for the machine learning model trained in the smart bracelet is X4, which is the data obtained by adding noise to the target data sample X0 four times. Furthermore, the machine learning model in the smart bracelet needs to process the input data once to obtain the output data. In the process of generating training samples, the smart bracelet performs three to four noise additions on the data sample X0 to obtain the data {X3, X4}, thereby constructing the training sample {X3, X4}. The smart bracelet trains the machine learning model based on the generated training sample {X3, X4} to obtain the trained machine model.

[0211] For a laptop computer, the input data of the machine learning model trained in the laptop computer is X3, that is, the data obtained after the target data sample X0 is denoised three times; and the machine learning model in the laptop computer needs to perform denoising on the input data three times to obtain the output data. In the process of generating training samples, the laptop computer performs denoising on the data sample X0 from 0 to 3 times to obtain the data {X0, X1, X2, X3}, thereby constructing the training sample {X0, X3}. The smart bracelet trains the machine learning model based on the generated training samples {X0, X3} to obtain the trained machine model. In addition to the smartphones, smart watches, smart bracelets and laptop computers shown in FIG8 , other terminal devices can also be used to train the machine learning model based on the data X T-5 The data between data X4 and data X4 is responsible for training the machine model, which is not shown one by one in Figure 8.

[0212] Optionally, during the process of generating training data for each terminal device, the terminal devices can send the generated training data to each other, thereby avoiding the need for each terminal device to independently and repeatedly generate the same training data, thereby improving the efficiency of generating training data. For example, for a smartphone, the smartphone needs to generate data {X T-3 , X T-2 , X T-1 , X T}, and smart watches also need to generate data X T-3 , so the smartphone can generate data X T-3 Send to the smart watch, thus eliminating the need to generate data X T-3 process.

[0213] In stage 3, multiple terminal devices send the trained machine learning models to the base station respectively.

[0214] Because each terminal device is responsible for different training content and uses different training data for machine learning models, the machine learning models trained on each terminal device are often different. After each terminal device completes one or more rounds of iterative model training and obtains a trained machine learning model, it sends the trained machine learning model to the base station, which allows the base station to aggregate the trained machine learning models.

[0215] In stage 4, the base station aggregates multiple trained machine learning models to obtain an aggregated model.

[0216] After the base station aggregates the trained machine learning models sent by each terminal device to obtain an aggregated model, it can continue to send the aggregated model to each terminal device, allowing each terminal device to continue iteratively training the aggregated model. Ultimately, after the machine learning model trained by the terminal device reaches the model training termination condition, the terminal device no longer trains the aggregated model sent by the base station. Instead, it uses the aggregated model sent by the base station last time as the model used for model inference, that is, it uses this aggregated model to perform subsequent data processing tasks.

[0217] The above describes the process of jointly training a machine learning model using multiple distributed devices. The following describes the process of jointly processing data using a machine learning model using multiple distributed devices. There are various scenarios for jointly processing data using a machine learning model using multiple distributed devices. The following details the joint data processing process in various scenarios, along with accompanying figures.

[0218] It should be noted that the multiple distributed devices that jointly process data through the machine learning model introduced in this embodiment can be multiple distributed devices that jointly train the machine learning model, that is, the multiple distributed devices first jointly train to obtain the machine learning model, and then jointly process the data through the same machine learning model. Alternatively, the multiple distributed devices in this embodiment can jointly process data through a preset machine learning model, that is, the multiple distributed devices do not perform the process of jointly training the machine learning model. In short, the two processes of the joint training process of the machine learning model and the joint data processing process can be integrated or independent, and this embodiment does not specifically limit this.

[0219] Please refer to Figure 9A, which is a flow chart of a data processing method 900 provided in an embodiment of the present application. As shown in Figure 9A, the data processing method 900 includes the following steps 901-905.

[0220] Step 901: Distributed device 1 determines data processing requirements.

[0221] In this embodiment, the data processing requirement in the distributed device 1 is to process the original data with noise to obtain the target data required by the distributed device 1. The quality of the target data is higher than the quality of the original data, that is, the noise in the target data is smaller than the noise in the original data. Generally speaking, the target data that the distributed device 1 expects to obtain can often be data used to perform other model training tasks. Therefore, the distributed device 1 can determine the data processing requirements based on the requirements of other model training tasks for the required data, and the data processing requirements are used to indicate the degree of processing of the original data. For example, the distributed device 1 expects to obtain high-quality image data so that it can subsequently train the image classification model based on the high-quality image data; therefore, the distributed device 1 can determine the degree of processing of the original image data based on the requirements of the image classification model for input data.

[0222] For example, when the original data is channel data, the target data that the distributed device 1 expects to obtain may be data used to train models such as a transmitter machine learning model, a receiver machine learning model, or a channel estimation machine learning model. For another example, when the original data is image data, the target data that the distributed device 1 expects to obtain may be data used to train models such as an image classification model, an image enhancement model, an image compression model, or an image detection model. Therefore, the distributed device 1 may determine the quality requirements of the target data based on the accuracy requirements of the model trained using the target data, and further determine the data processing requirements based on the quality gap between the target data and the original data.

[0223] Optionally, when a machine learning model (such as the diffusion model described above) is used to process the raw data, the data processing requirement may be the number of times the machine learning model is used to process the raw data. For example, if the quality of the required data is high, the distributed device 1 may determine the data processing requirement to be 10,000 times of sequential processing of the raw data; for another example, if the quality of the required data is not high, the distributed device 1 may determine the data processing requirement to be 1,000 times of sequential processing of the raw data.

[0224] In step 902, the distributed device 1 processes the original data through a machine learning model to obtain first data.

[0225] After determining the data processing requirements, distributed device 1 processes the raw data using a machine learning model based on its data processing capabilities to obtain first data. However, the data processing capabilities of distributed device 1 may not meet the data processing requirements of distributed device 1, and therefore the first data obtained by distributed device 1 is not the target data that distributed device 1 desires. The data processing capabilities of a distributed device may be related to the processing and storage resources on the distributed device, which is not specifically limited in this embodiment.

[0226] For example, assuming that the data processing requirement of distributed device 1 is to process the original data 1000 times using a machine learning model, and the data processing capacity of distributed device 1 only supports distributed device 1 to process the original data 200 times, distributed device 1 will process the original data 200 times using a machine learning model to obtain the first data. The first data still needs to be processed 800 times to obtain data that meets the data processing requirement of distributed device 1.

[0227] Alternatively, the distributed device 1 may process the raw data 200 times using a machine learning model. The distributed device 1 may reuse the machine learning models to obtain a master model composed of 200 machine learning models connected in sequence. The distributed device 1 may then input the raw data into the master model, and the 200 machine learning models in the master model may process the data in sequence to obtain the first data. The input of any machine learning model in the master model is the output of the previous machine learning model.

[0228] Alternatively, the process of distributed device 1 gradually processing the original data 200 times through the machine learning model can also mean: distributed device 1 processes the original data once through the machine learning model to obtain the first processed data; then, distributed device 1 inputs the first processed data into the machine learning model to obtain the second processed data; then, distributed device 1 continues to input the second processed data into the machine learning model to obtain the third processed data, and this cycle continues until distributed device 1 processes the data 200 times through the machine learning model to obtain the first data. In other words, distributed device 1 uses the data output by the machine learning model after each data processing as the input of the machine learning model in the next data processing process, thereby realizing multiple processing of data based on the same machine learning model.

[0229] Optionally, the machine learning model used to process raw data on the distributed device 1 is, for example, the diffusion model mentioned above.

[0230] Step 903 : Distributed device 1 sends first data to distributed device 2 .

[0231] Since the data processing capability of distributed device 1 cannot support distributed device 1 to complete data processing, that is, the first data processed by distributed device 1 does not meet the data processing requirements of distributed device 1, distributed device 1 sends the first data to distributed device 2 to request distributed device 2 to assist distributed device 1 in continuing to process the first data.

[0232] Optionally, while distributed device 1 sends the first data to distributed device 2, distributed device 1 may also send a first message to distributed device 2, where the first message is used to request distributed device 2 to process the first data. The first message may also indicate the number of times the first data is to be processed, that is, the number of times the first data is processed by distributed device 2. For example, assuming that the data processing requirement of distributed device 1 is to process the original data 1000 times step by step through a machine learning model, and distributed device 1 has only processed the original data 200 times step by step, the distributed device 1 may indicate in the first message that the number of times the first data is to be processed is 800 times. Optionally, the sending of the first data and the sending of the first message may be performed separately.

[0233] Optionally, the distributed device 1 may also negotiate with the distributed device 2 in advance so that the distributed device 2 can determine the number of times the data received from the distributed device 1 needs to be processed. In this case, the distributed device 1 only needs to send the first data to the distributed device 2, and does not need to send the above-mentioned first information to the distributed device 2 again. For example, in the case where the distributed device 1 needs to process a large amount of data step by step, the distributed device 1 negotiates with the distributed device 2 in advance so that the distributed device 2 can determine the number of times the received data needs to be processed. In this way, each time the distributed device 1 processes a piece of data a specific number of times, it sends the processed data to the distributed device 2, so that the distributed device 2 continues to process the data according to the content negotiated in advance, thereby avoiding the signaling overhead caused by the distributed device 1 repeatedly notifying the distributed device 2.

[0234] In this embodiment, the distributed device 1 and the distributed device 2 are, for example, the aforementioned terminal devices, or devices for implementing the functions of the aforementioned terminal devices. In a specific example, the distributed device 1 may be a smartwatch, and the distributed device 2 may be a smart phone.

[0235] In step 904, the distributed device 2 processes the first data through a machine learning model to obtain second data.

[0236] In this embodiment, the data processing capability of distributed device 2 can support distributed device 2 to assist distributed device 1 in completing data processing, that is, the second data processed by distributed device 2 can meet the data processing requirements of distributed device 1, and the second data is the data expected by distributed device 1.

[0237] For example, if distributed device 1 instructs distributed device 2 via first information on the number of times to process first data, distributed device 2 may progressively process the first data multiple times using the machine learning model based on the number of times indicated by the first information, thereby obtaining second data. For example, if distributed device 1 instructs distributed device 2 via the first information to process the first data 800 times, distributed device 2 may progressively process the first data 800 times using the machine learning model, thereby obtaining second data.

[0238] It should be noted that, in this embodiment, the data processing capability of the distributed device 2 can support the distributed device 2 to complete the processing of the first data for the number of times specified by the distributed device 1 .

[0239] Among them, the machine learning model used by distributed device 2 to process the first data can be the same as the machine learning model used by distributed device 1 to process the first data, so as to ensure the performance of distributed device 2 in noise reduction processing of the first data and ensure that the second data can meet the data processing requirements of distributed device 1.

[0240] Step 905 : Distributed device 2 sends second data to distributed device 1 .

[0241] After processing the second data, since the second data meets the data processing requirements of distributed device 1, distributed device 2 sends the second data to distributed device 1, thereby assisting distributed device 1 in processing data. In this way, after receiving the second data, distributed device 1 can perform other data processing tasks based on the second data, such as training other models based on the second data.

[0242] In this solution, by deploying machine learning models on different devices, data processing is completed jointly by the equipment of multiple devices to continuously improve the quality of the obtained data, reduce the data processing pressure of each device, and ensure that devices with weaker computing power can also obtain data of the required quality.

[0243] Please refer to Figure 9B, which is another flow chart of a data processing method 900 provided in an embodiment of the present application. As shown in Figure 9B, in another possible embodiment, the data processing method 900 may include the following steps 906-910. Among them, steps 906-910 are not sequentially associated with the above-mentioned steps 901-905, and steps 906-910 and the above-mentioned steps 901-905 may be two independent sets of steps. Distributed device 1 and distributed device 2 may complete the joint processing of data by executing the above-mentioned steps 901-905; distributed device 1 and distributed device 2 may also complete the joint processing of data by executing the above-mentioned steps 906-910.

[0244] Step 906: Distributed device 1 determines data processing requirements.

[0245] In this embodiment, step 906 is similar to the above-mentioned step 901. Please refer to the above-mentioned step 901 for details, and no further details will be given here.

[0246] Step 907: Distributed device 1 sends assistance request information to distributed device 2.

[0247] After determining the data processing requirements, distributed device 1 may determine whether its own data processing capabilities can meet the data processing requirements. If distributed device 1 determines that its own data processing capabilities cannot meet the data processing requirements, distributed device 1 may send a request for assistance to distributed device 2, requesting distributed device 2 to assist distributed device 1 in processing the data.

[0248] Optionally, the request for assistance message sent by distributed device 1 may indicate the number of times the original data is to be processed, that is, the number of times distributed device 2 processes the original data. For example, assuming that the data processing requirement of distributed device 1 is to process the original data 1000 times step by step using a machine learning model, and the data processing capacity of distributed device 1 only supports it to process the data 200 times step by step, distributed device 1 may indicate in the request for assistance message that the number of times the original data is to be processed is 800 times, that is, distributed device 1 instructs distributed device 2 to process the original data 800 times.

[0249] In step 908, the distributed device 2 processes the original data through a machine learning model to obtain first data.

[0250] After receiving the assistance request information sent by the distributed device 1, the distributed device 2 processes the original data through the machine learning model based on the instructions of the assistance request information to obtain the first data.

[0251] For example, assuming that the number of times the original data is to be processed indicated in the assistance request information is 800, the distributed device 2 processes the original data 800 times through the machine learning model to obtain the first data.

[0252] Step 909 : Distributed device 2 sends first data to distributed device 1 .

[0253] Optionally, while distributed device 2 is sending the first data to distributed device 1, distributed device 2 may also send instruction information to distributed device 1, indicating the number of times distributed device 2 processed the original data. For example, if distributed device 2 processes the original data 800 times using a machine learning model to obtain the first data, the instruction information sent by distributed device 2 indicates that the first data was obtained by processing the original data 800 times. It is understood that distributed device 2 may, based on its own capabilities, process the original data a number of times less than the number of pending processes indicated by distributed device 1.

[0254] Optionally, the sending of the first data and the sending of the indication information may be performed separately.

[0255] In step 910 , the distributed device 1 processes the first data through a machine learning model to obtain second data.

[0256] After receiving the first data, the distributed device 1 can determine the number of times the distributed device 1 processes the first data based on the data processing requirements of the device and the number of times the distributed device 2 processes the original data, thereby processing the first data through a machine learning model to obtain the second data.

[0257] For example, if distributed device 1 instructs distributed device 2 to process the original data 800 times through a request for assistance, distributed device 1 can process the original data 1000 times based on its own data processing requirements and determine that it needs to process the received first data 200 times. Therefore, distributed device 1 continues to process the first data 800 times using the machine learning model to obtain the second data.

[0258] In the above-described data processing method 900, after distributed device 1 sends first data or requests assistance information to distributed device 2, distributed device 2 can assist distributed device 1 in completing data processing. However, in some scenarios, such as when distributed device 1 has a high data processing requirement or when distributed device 2 has a low data processing capability, distributed device 2 may find it difficult to assist distributed device 1 in completing data processing on its own. Therefore, distributed device 2 may also send the processed data to other distributed devices, allowing them to continue assisting in completing data processing. Alternatively, distributed device 1 may also request other distributed devices to continue assisting in processing, which is similar to requesting assistance from distributed device 2 and will not be further described here.

[0259] Please refer to Figure 10A, which is a flow chart of a data processing method 1000 provided in an embodiment of the present application. As shown in Figure 10A, the data processing method 1000 includes the following steps 1001-1007.

[0260] Step 1001: Distributed device 1 determines data processing requirements.

[0261] In step 1002, the distributed device 1 processes the original data through a machine learning model to obtain first data.

[0262] Step 1003 : Distributed device 1 sends first data to distributed device 2 .

[0263] In this embodiment, steps 1001-1003 are similar to the above-mentioned steps 901-903. Please refer to the above-mentioned steps 901-903 for details, and no further details will be given here.

[0264] In step 1004 , the distributed device 2 processes the first data through a machine learning model to obtain second data.

[0265] In this embodiment, the data processing capabilities of distributed device 2 are insufficient to support distributed device 2 assisting distributed device 1 in completing data processing. For example, suppose distributed device 1's data processing requirement is to process raw data 1,000 times using a machine learning model, but distributed device 1 has only processed the raw data 200 times. Therefore, distributed device 1 may instruct distributed device 2 to process the first data 800 times. However, the data processing capabilities of distributed device 2 are insufficient to support distributed device 2 processing the first data 800 times. Therefore, distributed device 2 may only be able to process the first data 200 times to obtain the second data.

[0266] That is to say, the second data obtained by the distributed device 2 after processing the first data through the machine learning model still does not meet the data processing requirements of the distributed device 1, that is, the second data is not the target data expected by the distributed device 1.

[0267] Step 1005 : Distributed device 2 sends second data to distributed device 3 .

[0268] Since the second data obtained after the distributed device 2 processes the first data is not the target data expected by the distributed device 1, the distributed device 2 may continue to request other distributed devices to assist in completing the data processing.

[0269] Specifically, distributed device 2 sends the second data to distributed device 3, so that distributed device 3 continues to process the second data, thereby assisting distributed device 1 in completing data processing.

[0270] Optionally, while distributed device 2 is transmitting the second data to distributed device 3, distributed device 2 may also transmit second information to distributed device 3, where the second information indicates the number of times the second data transmitted by distributed device 2 is pending for processing. The number of times the second data is pending for processing may be calculated based on the number of times the first data is pending for processing indicated in the first information and the number of times distributed device 2 actually processes the first data. Optionally, the transmission of the second data and the transmission of the second information may be performed separately.

[0271] For example, suppose that the data processing requirement of distributed device 1 is to gradually process the original data 1,000 times through a machine learning model, and distributed device 1 only gradually processes the original data 200 times, so distributed device 1 indicates through the first information that the number of times the first data is to be processed is 800 times; after distributed device 2 receives the first data and the first information, distributed device 2 processes the first data 200 times to obtain the second data; therefore, distributed device 2 can send the second data and the second information to distributed device 3, and the second information is used to indicate that the number of times the second data is to be processed is 600 times (800-200).

[0272] Step 1006 , distributed devices 3 to N assist in processing data in sequence.

[0273] Similarly, after receiving the second data sent by distributed device 2, distributed device 3 continues to process the second data. Furthermore, if the number of times distributed device 3 has processed the second data is still less than the number of times the second data needs to be processed, distributed device 3 continues to send the processed data to the next distributed device after distributed device 3, instructing the subsequent distributed devices to continue assisting in completing the data processing, until distributed device N processes the target data that meets the data processing requirements of distributed device 1.

[0274] In this embodiment, the distributed device 3 and the distributed device N may be the same distributed device or different distributed devices. In FIG10A , the distributed device 3 and the distributed device N are drawn as different distributed devices.

[0275] Step 1007 : Distributed device N sends target data to distributed device 1 .

[0276] After processing the target data required by distributed device 1, distributed device N sends the target data to distributed device 1, as the target data meets the data processing requirements of distributed device 1, thereby assisting distributed device 1 in processing data. In this way, after receiving the target data, distributed device 1 can perform other data processing tasks based on the target data, such as training other models based on the second data.

[0277] Optionally, distributed device N can send target data directly to distributed device 1; distributed device N can also send target data to the previous distributed device of distributed device N (i.e., distributed device N-1 that requests distributed device N to assist in processing data) so that the target data can be sent to distributed device 1 hop by hop.

[0278] Please refer to Figure 10B, which is another flow chart of a data processing method 1000 provided in an embodiment of the present application. As shown in Figure 10B, in another possible embodiment, the data processing method 1000 may include the following steps 1008-1014. Among them, steps 1008-1014 are not sequentially associated with the above-mentioned steps 1001-1007, and steps 1008-1014 and the above-mentioned steps 1001-1007 may be two independent sets of steps. Distributed device 1-distributed device N can complete the joint processing of data by executing the above-mentioned steps 1001-1007; distributed device 1-distributed device N can also complete the joint processing of data by executing the above-mentioned steps 1008-1014.

[0279] Step 1008: The distributed device 1 determines the data processing requirements.

[0280] Step 1009 : Distributed device 1 sends assistance request information to distributed device 2 .

[0281] In this embodiment, steps 1008-1009 are similar to the above-mentioned steps 906-907. Please refer to the above-mentioned steps 906-907 for details, and no further details will be given here.

[0282] In step 1010, the distributed device 2 processes the original data through a machine learning model to obtain intermediate data.

[0283] Unlike step 908 above, distributed device 2's data processing capabilities are insufficient to support distributed device 2 completing the number of data processing operations indicated by distributed device 1 in the request for assistance. Therefore, distributed device 2 processes the raw data using a machine learning model based on its own data processing capabilities to generate intermediate data. The number of data processing operations corresponding to the intermediate data is less than the number of data processing operations indicated by distributed device 1 in the request for assistance.

[0284] For example, suppose that distributed device 1 indicates in the request for assistance information that the number of times the original data needs to be processed is 800 times, and the data processing capability of distributed device 2 only supports it to process the original data 300 times, then the intermediate data obtained after the distributed device processes the original data 300 times through the machine learning model cannot meet the needs of distributed device 1, that is, the intermediate data is not the data expected by distributed device 1.

[0285] Step 1011 : Distributed device 2 sends intermediate data to distributed device 3 .

[0286] Since the intermediate data obtained after the distributed device 2 processes the original data is not the data expected by the distributed device 1, the distributed device 2 may continue to request other distributed devices to assist in completing the data processing.

[0287] Specifically, distributed device 2 sends the intermediate data to distributed device 3, so that distributed device 3 continues to process the intermediate data, thereby assisting distributed device 1 in completing the data processing.

[0288] Optionally, while distributed device 2 is sending the intermediate data to distributed device 3, distributed device 2 may also send indication information to distributed device 3, indicating the number of times the intermediate data sent by distributed device 2 is pending for processing. The number of times the intermediate data is pending for processing may be calculated based on the number of times the original data is pending for processing indicated in the assistance request information and the number of times distributed device 2 actually processes the original data. Optionally, the sending of the intermediate data and the sending of the indication information may be performed separately.

[0289] In step 1012 , distributed devices 3 to N assist in processing data in sequence to obtain the first data required by distributed device 1 .

[0290] Similarly, after receiving the intermediate data sent by distributed device 2, distributed device 3 continues to process the intermediate data. Furthermore, if the number of times distributed device 3 has processed the intermediate data is still less than the number of times the intermediate data needs to be processed, distributed device 3 continues to send the processed data to the next distributed device after distributed device 3, instructing the subsequent distributed devices to continue assisting in completing the data processing, until distributed device N processes the first data that meets the data processing requirements of distributed device 1.

[0291] Step 1013 : Distributed device N sends first data to distributed device 1 .

[0292] The first data can satisfy the data processing requirements indicated by the distributed device 1 in the assistance request information, so the distributed device N sends the first data to the distributed device 1 .

[0293] In step 1014, the distributed device 1 processes the first data through a machine learning model to obtain second data.

[0294] Methods 900 and 1000 above describe the process of a distributed device requesting assistance from other distributed devices to complete data processing. In some scenarios, different distributed devices may need to process the same type of data, but their data processing requirements may differ. In such cases, a central device can coordinate the data processing needs of each distributed device, thereby enabling joint data processing among the different distributed devices.

[0295] Please refer to Figure 11, which is a flow chart of a data processing method 1100 provided in an embodiment of the present application. As shown in Figure 11, the data processing method 1100 includes the following steps 1101-1108.

[0296] Step 1101: Multiple distributed devices send data processing requirements to a central device respectively.

[0297] In this embodiment, three distributed devices are used as an example. For example, the multiple distributed devices include distributed device 1, distributed device 2, and distributed device 3. The data required by the multiple distributed devices is of the same type, but different distributed devices have different quality requirements for the required data, that is, different distributed devices require different numbers of times to perform noise reduction processing on the original data.

[0298] For example, assuming that the data required by multiple distributed devices are all image data, distributed device 1 needs to use image data to train an image classification model. The image classification model does not have high requirements on the quality of image data. Therefore, the data processing requirements of distributed device 1 can specifically be to perform noise reduction processing on the original image data 1000 times.

[0299] The distributed device 2 may require image data to train a semantic segmentation model, which is used to identify objects in the image. Therefore, the semantic segmentation model has high requirements on the quality of image data; the data processing requirements of the distributed device 2 may specifically be to perform denoising processing on the original image data 5000 times.

[0300] Furthermore, distributed device 3 may require image data to train an image enhancement model. This image enhancement model is used to identify specific objects in an image and enhance the clarity of the identified specific objects. Therefore, the image enhancement model has the highest quality requirements for image data. Specifically, the data processing requirement of distributed device 3 may be to perform noise reduction processing on the original image data 10,000 times.

[0301] Step 1102: The central device determines the order in which each distributed device processes data.

[0302] Since the data processing requirements of each distributed device are different, the central device can determine the order in which each distributed device processes data based on the number of data processing times required by each distributed device. For example, the central device first determines the number of data processing times in the data processing requirements of each distributed device, and then determines the order in which each distributed device processes data in ascending order of the number of data processing times. The smaller the number of data processing times in the data processing requirements of a distributed device, the earlier the distributed device processes data; the larger the number of data processing times in the data processing requirements of a distributed device, the later the distributed device processes data.

[0303] For example, assuming that the data processing requirement of distributed device 1 is to perform noise reduction processing on the original image data 1000 times, the data processing requirement of distributed device 2 is to perform noise reduction processing on the original image data 5000 times, and the data processing requirement of distributed device 3 is to perform noise reduction processing on the original image data 10000 times, then the order in which the three distributed devices process data is: distributed device 1 → distributed device 2 → distributed device 3.

[0304] Step 1103: The central device sends instruction information to each distributed device to instruct each distributed device on the order in which to process data.

[0305] After determining the order in which each distributed device processes data, the central device sends instructions to each distributed device, indicating the order in which each distributed device processes data. This allows each distributed device to determine which distributed device to receive processed data from and which distributed device to send processed data to after receiving the instructions from the central device.

[0306] Optionally, when the ability to process data on each distributed device is relatively stable, that is, when the computing resources and storage resources allocated to data processing on each distributed device are relatively stable, the central device can also indicate the number of times each distributed device needs to process data in the instruction information sent to each distributed device.

[0307] In one possible example, if distributed device 1, distributed device 2, and distributed device 3 require denoising the original image data 1000 times, 5000 times, and 10,000 times, respectively, the instruction message 1 sent by the central device to distributed device 1 may specifically include: the previous hop node is empty, the number of local data processing times is 1000, and the next hop node is distributed device 2. This means that distributed device 1 is the first node to begin processing data, and distributed device 1 needs to process the data 1000 times using a machine learning model and send the processed data to distributed device 2. Furthermore, the instruction message 2 sent by the central device to distributed device 2 may specifically include: the previous hop node is distributed device 1, the number of local data processing times is 4000 (i.e., 5000-1000), and the next hop node is distributed device 3. The instruction message 3 sent by the central device to distributed device 3 may specifically include: the previous hop node is distributed device 2, the number of local data processing times is 5000, and the next hop node is empty. In this example, the data processing capabilities of each distributed device meet the data processing requirements of this device. Therefore, after any distributed device receives data that has been processed a certain number of times from other distributed devices, it can continue to process the processed data to obtain data that meets the data processing requirements of this device.

[0308] In other cases, the data processing capabilities of distributed devices may not be able to meet their own data processing needs. If the central device continues to determine the method for distributed devices to jointly process data based on the data processing times in the data processing requirements of each distributed device, some distributed devices may be unable to complete data processing. Therefore, in this example, the central device can determine the method for distributed devices to jointly process data based on the data processing times in the data processing requirements of each distributed device and the data processing capabilities of each distributed device.

[0309] For example, if the data processing requirements of distributed device 1, distributed device 2, and distributed device 3 are to perform noise reduction processing on the original image data 1000 times, 5000 times, and 10000 times, respectively, and the data processing capability of distributed device 1 supports it performing noise reduction processing on the data 1000 times, the data processing capability of distributed device 2 supports it performing noise reduction processing on the data 2000 times, and the data processing capability of distributed device 3 supports it performing noise reduction processing on the data 9000 times, the instruction information 1 sent by the central device to distributed device 1 may specifically be: the previous hop node is empty, the number of local data processing times is 1000, and the next hop node is distributed device 2. The instruction information 2 sent by the central device to distributed device 2 may specifically be: the previous hop node is distributed device 1, the number of local data processing times is 2000, and the next hop node is distributed device 3; the previous hop node is distributed device 3, the number of local data processing times is 0, and the next hop node is empty. The instruction information 3 sent by the central device to the distributed device 3 may specifically be: the previous hop node is the distributed device 2, the next hop node is the distributed device 2 when the local data processing times is 2000, and the next hop node is empty when the local data processing times is 7000.

[0310] In step 1104, the distributed device 1 processes the original data T1 times through the machine learning model to obtain first data.

[0311] After receiving the instruction information sent by the central device, the distributed device 1 can determine that it is the first device to process the data. Therefore, the distributed device 1 processes the original data T1 times through the machine learning model to obtain the first data.

[0312] Optionally, if the data processing capabilities of each distributed device are relatively stable, the central device can specify the number of times each distributed device processes data through instruction information. The number of times distributed device 1 processes raw data can match the data processing requirements of distributed device 1. That is, if the data processing requirement of distributed device 1 is to process raw data T1 times, the number of times distributed device 1 actually processes raw data is also T1.

[0313] Optionally, when the ability to process data on each distributed device fluctuates, the central device does not specify the number of times each distributed device processes data. The number of times the distributed device 1 processes the original data may not match the data processing requirements of the distributed device 1. That is, the number of times the distributed device 1 actually processes the original data may be greater than or less than the number of data processing times required by the distributed device 1. For example, the data processing requirement of the distributed device 1 is to process the original data N1 times, and the number of times the distributed device 1 actually processes the original data is T1 times, where N1 may be greater than or less than T1. When the computing resources and storage resources on the distributed device 1 are relatively abundant, the number of times the distributed device 1 actually processes the original data T1 may be greater than the required number of data processing times N1; when the computing resources or storage resources on the distributed device 1 are relatively tight, the number of times the distributed device 1 actually processes the original data T1 may be less than the required number of data processing times N1.

[0314] Step 1105 : Distributed device 1 sends first data to distributed device 2 .

[0315] In this embodiment, the instruction information received by distributed device 1 from the central device also indicates that distributed device 1 needs to send processed data to distributed device 2. Therefore, after distributed device 1 processes the original data and obtains the first data, distributed device 1 sends the first data to distributed device 2.

[0316] Optionally, when the central device does not indicate the number of times each distributed device needs to process the data, distributed device 1 may send information to distributed device 2 to indicate the number of times distributed device 1 has processed the original data.

[0317] In step 1106, the distributed device 2 processes the first data T2 times using a machine learning model to obtain second data.

[0318] After receiving the instruction information sent by the central device, distributed device 2 can determine that it needs to receive data from distributed device 1 and continue to process the received data. Therefore, distributed device 2 processes the received first data T2 times using the machine learning model to obtain second data. In particular, if the second data obtained by distributed device 2 processing the first data T2 times can meet the data processing requirements of distributed device 2, the second data is the data required by distributed device 2.

[0319] Optionally, if the data processing capabilities of each distributed device are relatively stable, the central device can specify the number of times distributed device 2 is to process data through instruction information. After distributed device 2 processes the first data T2 times according to the instruction of the central device, it can obtain the data required by distributed device 2. For example, assuming that the data processing requirement of distributed device 2 is to process the original data T1+T2 times, since the first data received by distributed device 2 is the data obtained after the original data is processed T1 times, the second data obtained by distributed device 2 after processing the first data T2 times according to the instruction of the central device is the data required by distributed device 2.

[0320] Optionally, when the data processing capabilities of each distributed device fluctuate, the number of times distributed device 2 processes the first data may be determined by the number of data processing times corresponding to the first data and the data processing requirements of distributed device 2. For example, when distributed device 1 indicates to distributed device 2 that distributed device 1 actually processed the original data T1 times, distributed device 2 may process the original data N2 times based on its own data processing requirements, and determine that the number of times distributed device 2 needs to process the first data may be N2-T1=T2 times.

[0321] Alternatively, if the data processing capability of distributed device 2 only supports processing the first data S1 times (S1 < T2), i.e., distributed device 2 does not support processing the first data T2 times, distributed device 2 may process the first data S1 times to obtain second data. Wherein, the second data is not the data required by distributed device 2, distributed device 2 may send the second data to other distributed devices to request the other distributed devices to assist distributed device 2 in continuing to process the second data.

[0322] Alternatively, if the data processing capability of distributed device 2 supports processing the first data S2 times (S2>T2), that is, if distributed device 2 supports processing the first data more than T2 times, distributed device 2 may process the first data S2 times to obtain the second data. In the process where distributed device 2 can process the first data S2 times, the data obtained when distributed device 2 processes the first data T2 times is the data required by distributed device 2.

[0323] Step 1107 : Distributed device 2 sends second data to distributed device 3 .

[0324] In this embodiment, the instruction information received by distributed device 2 from the central device also indicates that distributed device 2 needs to send processed data to distributed device 3. Therefore, after distributed device 2 processes the first data and obtains the second data, distributed device 2 sends the second data to distributed device 3.

[0325] In step 1108, the distributed device 3 processes the second data T3 times through the machine learning model to obtain third data.

[0326] After receiving the instruction information sent by the central device, distributed device 3 can determine that it needs to receive data from distributed device 2 and continue to process the received data. Therefore, distributed device 3 processes the second data T3 times using the machine learning model to obtain third data. In particular, if the third data obtained by distributed device 3 processing the first data T3 times can meet the data processing requirements of distributed device 3, the third data is the data required by distributed device 3.

[0327] Among them, the process of distributed device 3 processing the second data in various situations is similar to the process of distributed device 2 processing the first data. For details, please refer to the process of distributed device 2 processing the first data in the above step 1106, which will not be repeated here.

[0328] It can be understood that the above method 1100 is described by taking three distributed devices jointly processing data as an example. In actual applications, two or more distributed devices can jointly process data based on the above process. This embodiment does not limit the number of distributed devices that jointly process data.

[0329] Furthermore, the above method 1100 describes how to coordinate data processing among distributed devices when their data processing requirements differ. In some special scenarios, some distributed devices may have the same data processing requirements. The central device can then allocate data processing tasks to these distributed devices based on their data processing capabilities.

[0330] Please refer to Figure 12, which is a flow chart of a data processing method 1200 provided in an embodiment of the present application. As shown in Figure 12, the data processing method 1200 includes the following steps 1201-1207.

[0331] In step 1201 , distributed device 1 and distributed device 2 respectively send data processing requirements and data processing capabilities to the central device.

[0332] In this embodiment, the data processing requirements of distributed device 1 and distributed device 2 are the same. For example, the data processing requirements of distributed device 1 and distributed device 2 are both to perform noise reduction processing on the original data 1000 times.

[0333] In addition, the data processing capabilities of the distributed device 1 and the distributed device 2 may be the same or different, and this embodiment does not specifically limit this.

[0334] In step 1202, the central device determines the order and number of times each distributed device processes data.

[0335] In this embodiment, since the data processing requirements of distributed device 1 and distributed device 2 are the same, which means that the data required by distributed device 1 and the data required by distributed device 2 are the same, the central device can allocate part of the data processing process to distributed device 1 and allocate another part of the data processing process to distributed device 2.

[0336] Optionally, the central device can determine the order in which distributed devices process data in a variety of ways. For example, the central device can randomly determine the order in which distributed devices 1 and 2 process data. Alternatively, the central device can determine the order in which distributed devices 1 and 2 process data based on the source of the data to be processed by distributed devices 1 and 2. For example, assuming that the data to be processed by distributed devices 1 and 2 originates from distributed device 0, and distributed device 1 is located between distributed devices 0 and 2, the central device can determine that distributed device 1 processes the data first, followed by distributed device 2 continuing to process the data processed by distributed device 1. Alternatively, the central device can determine the order in which distributed devices 1 and 2 process data based on the next-hop nodes of the data processed by distributed devices 1 and 2. For example, assuming that the data ultimately processed by distributed devices 1 and 2 needs to be sent to distributed device 3, and distributed device 1 is located between distributed devices 2 and 3, the central device can determine that distributed device 2 processes the data first, followed by distributed device 1 continuing to process the data processed by distributed device 2, so that distributed device 1 can send the ultimately processed data to distributed device 3 more quickly.

[0337] Optionally, the number of times each distributed device processes data may be determined by the central device based on the data processing capabilities of each distributed device. The higher the data processing capabilities of a distributed device, the greater the number of times the central device may determine that the distributed device processes data; and the lower the data processing capabilities of a distributed device, the less frequent the number of times the central device may determine that the distributed device processes data.

[0338] In step 1203, the central device sends instruction information to each distributed device to instruct each distributed device on the order and number of times to process data.

[0339] For example, if distributed device 1 has a lower data processing capability and distributed device 2 has a higher data processing capability, the central device may send instruction information 1 to distributed device 1, instructing distributed device 1 to process the data 200 times and send the processed data to distributed device 2. In addition, the central device may also send instruction information 2 to distributed device 2, instructing distributed device 2 to receive data from distributed device 1, process the received data 800 times, and send the processed data to distributed device 1.

[0340] In step 1204, the distributed device 1 processes the data to be processed N1 times through the machine learning model to obtain first data.

[0341] Based on the instructions of the central device, the distributed device 1 performs N1 data processing on the data to be processed using the machine learning model. The data to be processed can be raw data pre-stored on the distributed device 1; the data to be processed can also be processed data sent to the distributed device 1 by other distributed devices. This embodiment does not specifically limit the data to be processed.

[0342] Step 1205 : Distributed device 1 sends first data to distributed device 2 .

[0343] In step 1206, the distributed device 2 processes the first data N2 times through the machine learning model to obtain second data.

[0344] In this embodiment, the data processing requirements of distributed device 1 and distributed device 2 are to process the data to be processed N1+N2 times. Therefore, after distributed device 2 processes the first data N2 times, the second data obtained is the data required by distributed device 1 and distributed device 2. At this time, distributed device 2 can use the obtained second data to perform other tasks, such as using the second data to train an image processing model.

[0345] Step 1207 : Distributed device 2 sends second data to distributed device 1 .

[0346] Since the central device instructs the distributed device 2 to send the processed data to the distributed device 1, the distributed device 2 sends the second data to the distributed device 1 after obtaining the second data, so that the distributed device 1 performs other tasks based on the second data.

[0347] In order to implement the above-mentioned method embodiment, the present application also provides a communication device. Please refer to Figure 13. The embodiment of the present application provides a communication device 1300. The communication device 1300 can implement the functions of the terminal device (or network device) in the above-mentioned method embodiment, and thus can also achieve the beneficial effects of the above-mentioned method embodiment. In the embodiment of the present application, the communication device 1300 can be a terminal device (or network device), or it can be an integrated circuit or component inside the terminal device (or network device), such as a chip. The following embodiments are described by taking the communication device 1300 as a terminal device or a network device as an example.

[0348] In a possible embodiment, the communication device 1300 includes: a transceiver module 1301 and a processing module 1302 .

[0349] The transceiver module 1301 is used to receive first data from the second device, where the first data is data processed by a first machine learning model; the processing module 1302 is used to process the first data through a second machine learning model to obtain second data. The structure of the first machine learning model is the same as that of the second machine learning model. The communication device and the second device are used to jointly execute data processing.

[0350] In one possible implementation, the second machine learning model is a diffusion model, and the second machine learning model is used to perform noise reduction processing on the first data.

[0351] In a possible implementation, the transceiver module 1301 is further configured to receive first information from the second device, where the first information is used to request the communication device to process the first data.

[0352] In one possible implementation, the first information is used to indicate that the number of times the first data is to be processed is the first number; the processing module 1302 is also used to process the first data the first number through a second machine learning model to obtain second data, wherein the capability of the first device supports the completion of the first number of processing of the first data.

[0353] In one possible implementation, the transceiver module 1301 is further used to send the second data to the second device; or, the transceiver module 1301 is used to send the second data to the source device, wherein the first information is also used to indicate information of the source device, and the source device is the first device to request assistance in processing data.

[0354] In one possible implementation, the first information is used to indicate that the number of times the first data is to be processed is the first number; the processing module 1302 is also used to process the first data for the second number through the second machine learning model to obtain second data, wherein the first number is greater than the second number, and the capability of the first device does not support the completion of the first number of processing of the first data; the transceiver module 1301 is also used to send the second data and the second information to the third device; wherein the second information is used to indicate that the number of times the second data is to be processed is the third number, and the third number is the difference between the first number and the second number, and the third device is used to assist the communication device in performing data processing.

[0355] In a possible implementation, the transceiver module 1301 is further configured to send assistance request information to the second device, where the assistance request information is used to request the second device to assist in processing data.

[0356] In one possible implementation, the transceiver module 1301 is further used to send third information to the central device, where the third information is used to indicate the number of data processing times required by the communication device; the transceiver module 1301 is further used to receive feedback information from the central device, where the feedback information is used to indicate that the second device is an assisting node.

[0357] In one possible implementation, the transceiver module 1301 is further used to receive fourth information from the central device, where the fourth information is used to indicate the number of times the communication device needs to perform processing on the data received from the second device; the processing module 1302 is further used to process the first data through a second machine learning model based on the fourth information to obtain the second data required by the communication device.

[0358] In a possible implementation, the fourth information is further used to indicate information of a third device, where the third device is a device to receive processed data; the transceiver module 1301 is further used to send the second data to the third device according to the fourth information.

[0359] In one possible implementation, the transceiver module 1301 is further used to receive fifth information from the second device, where the fifth information is used to indicate the number of times the first data has been processed; the processing module 1302 is further used to process the first data through a second machine learning model based on the number of processing times and the number of processing times of the data required by the communication device to obtain the second data required by the communication device.

[0360] In another possible embodiment, the processing module 1302 is used to process the original data through a first machine learning model to obtain first data; the transceiver module 1301 is used to send the first data to the second device to request the second device to assist in processing the first data; the transceiver module 1301 is used to receive second data sent by the second device or other devices, the second data is obtained by processing the first data, the second data is obtained based on the second machine learning model, and the structure of the first machine learning model is the same as the structure of the second machine learning model.

[0361] In one possible implementation, the first machine learning model is a diffusion model, and the first machine learning model is used to perform noise reduction processing on the original data.

[0362] In one possible implementation, the transceiver module is also used to send first information to the second device, where the first information is used to request the second device to perform processing on the first data, and the first information is also used to indicate the number of times the first data is to be processed, which is determined based on the number of times the original data needs to be processed and the number of times the first device performs processing on the original data.

[0363] In another possible embodiment, the transceiver module 1301 is used to receive first information from a first device and second information from a second device, where the first information is used to indicate a first processing number of times the data is required by the first device, and the second information is used to indicate a second processing number of times the data is required by the second device, and the data processing model corresponding to the first processing number is the same as the data processing model corresponding to the second processing number; the transceiver module 1301 is used to send third information to the second device, where the third information is used to instruct the second device to send processed data to the first device, where the second processing number of times the data is required by the second device is less than or equal to the first processing number of times the data is required by the first device.

[0364] In a possible implementation, the transceiver module 1301 is further configured to send fourth information to the first device, where the fourth information is configured to indicate a number of times the first device needs to process data received from the second device.

[0365] Optionally, when the communication device 1300 is a terminal device or a network device, the transceiver module 1301 in the communication device 1300 may be a transceiver, and the processing module 1302 may be a processor. When the communication device 1300 is an integrated circuit or component within the terminal device or the network device, for example, when the communication device 1300 is a chip, the transceiver module 1301 in the communication device 1300 may be the output and input pins on the chip, and the processing module 1302 may be the computing component on the chip. For another example, when the communication device 1300 is a chip system, the transceiver module 1301 in the communication device 1300 may be a communication interface on the chip system, and the processing module 1302 may be a processing core on the chip system.

[0366] Please refer to Figure 14. An embodiment of the present application provides a model training device 1400. The model training device 1400 can implement the functions of the terminal device (or network device) in the above method embodiment, and thus can also achieve the beneficial effects of the above method embodiment. In the embodiment of the present application, the model training device 1400 can be a terminal device (or network device), or it can be an integrated circuit or component inside the terminal device (or network device), such as a chip. The following embodiments are explained using the model training device 1400 as a terminal device or a network device as an example.

[0367] As shown in Figure 14, the model training device 1400 includes: a transceiver module 1401 and a processing module 1402. The transceiver module 1401 is used to obtain a training sample set, where the training sample set includes first data and second data, where the first data is obtained based on the second data, and the second data is a training label for the first data; the processing module 1402 is used to train a first machine learning model based on the training sample set to obtain a trained first machine learning model, where the first machine learning model is used to process the first data; the transceiver module 1401 is used to send the trained first machine learning model to a second device, where the second device is used to aggregate machine learning model devices with the same structure and different parameters obtained by training multiple devices.

[0368] In one possible implementation, the transceiver module 1401 is also used to send first information to a third device, where the first information is used to indicate capabilities related to model training on the model training device, and the third device is used to determine the training content that multiple devices are responsible for based on the capabilities of multiple devices participating in the machine learning model training; the transceiver module 1401 is also used to receive second information from the third device, where the second information is used to indicate the number of times the first machine learning model trained on the model training device processes the input data, and the second information is also used to indicate the input data requirements of the first machine learning model; the processing module 1402 is also used to: process the original data according to the input data requirements indicated by the second information and the number of times the first machine learning model processes the input data to obtain second data; and process the second data according to the number of times the first machine learning model processes the input data indicated by the second information to obtain first data.

[0369] In one possible implementation, the transceiver module 1401 is also used to receive a second machine learning model from a second device; the processing module 1402 is also used to train the second machine learning model based on the training sample set to obtain a trained second machine learning model; the transceiver module 1401 is also used to send the trained second machine learning model to the second device.

[0370] In another possible embodiment, the transceiver module 1401 is used to receive multiple capability information, where the multiple capability information comes from multiple different devices, and each capability information in the multiple capability information is used to indicate the capability related to model training on the device; the sending module 1403 is used to send different training configuration information to multiple different devices according to the multiple capability information, where the training configuration information is used to indicate the number of times the machine learning model trained on the device processes the input data. The training configuration information is also used to indicate the input data requirements of the machine learning model trained on the device, and the machine learning models trained by multiple different devices are models with the same structure.

[0371] Optionally, when the above-mentioned model training device 1400 is a terminal device or a network device, the transceiver module 1401 in the model training device 1400 may be a transceiver, and the processing module 1402 may be a processor. In the case where the model training device 1400 is an integrated circuit or component inside the terminal device or the network device, for example, when the model training device 1400 is a chip, the transceiver module 1401 in the model training device 1400 may be the output and input pins on the chip, and the processing module 1402 may be the computing component on the chip. For another example, when the model training device 1400 is a chip system, the transceiver module 1401 in the model training device 1400 may be the communication interface on the chip system, and the processing module 1402 may be the processing core on the chip system.

[0372] Please refer to Fig. 15, which is another schematic structural diagram of a communication device 1500 provided in this application. The communication device 1500 at least includes an input and output interface 1502. The communication device 1500 may be a chip or an integrated circuit.

[0373] Optionally, the communication device further includes a logic circuit 1501 .

[0374] The transceiver module 1301 shown in FIG13 may be a communication interface, which may be the input / output interface 1502 in FIG15 , which may include an input interface and an output interface. Alternatively, the communication interface may be a transceiver circuit, which may include an input interface circuit and an output interface circuit.

[0375] Optionally, the input / output interface 1502 is configured to obtain AI model information of the first network device; and the logic circuit 1501 is configured to determine AI performance information of the first network device based on the AI ​​model information of the first network device. The logic circuit 1501 and the input / output interface 1502 may also perform other steps performed by the terminal device in any of the aforementioned embodiments and achieve corresponding beneficial effects, which will not be further described here.

[0376] Optionally, the logic circuit 1501 is used to generate AI model information of the first network device; the input / output interface 1502 is used to send the AI ​​model information of the first network device. The logic circuit 1501 and the input / output interface 1502 may also perform other steps performed by the network device in any embodiment and achieve corresponding beneficial effects, which will not be repeated here.

[0377] In a possible implementation, the processing module 1302 shown in FIG. 13 may be the logic circuit 1501 in FIG. 15 .

[0378] Optionally, the logic circuit 1501 may be a processing device, and the functions of the processing device may be partially or entirely implemented by software. The functions of the processing device may be partially or entirely implemented by software.

[0379] Optionally, the processing device 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 corresponding processing and / or steps in any one of the method embodiments.

[0380] Alternatively, the processing device may include only a processor. A memory for storing the computer program is located outside the processing device, and the processor is connected to the memory via circuits / wires to read and execute the computer program stored in the memory. The memory and processor may be integrated or physically separate.

[0381] 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 processor units (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.

[0382] Please refer to Figure 16, which shows the communication device 1600 involved in the above-mentioned embodiments provided in an embodiment of the present application. The communication device 1600 can specifically be a communication device serving as a terminal device in the above-mentioned embodiments. The example shown in Figure 16 is that the terminal device is implemented through the terminal device (or a component in the terminal device).

[0383] Herein, a possible logical structure diagram of the communication device 1600 is shown. The communication device 1600 may include but is not limited to at least one processor 1601 and a communication port 1602 .

[0384] Further optionally, the device may also include at least one of a memory 1603 and a bus 1604 . In an embodiment of the present application, the at least one processor 1601 is used to control and process the actions of the communication device 1600 .

[0385] In addition, processor 1601 can be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field-programmable gate array or other programmable logic device, a transistor logic device, a hardware component, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. The processor can also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a digital signal processor and a microprocessor, and so on. Those skilled in the art will clearly understand that for the sake of convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0386] It should be noted that the communication device 1600 shown in Figure 16 can be specifically used to implement the steps implemented by the terminal device in the aforementioned method embodiment and achieve the corresponding technical effects of the terminal device. The specific implementation methods of the communication device shown in Figure 16 can refer to the description in the aforementioned method embodiment and will not be repeated here.

[0387] Please refer to Figure 17, which is a structural diagram of the communication device 1700 involved in the above-mentioned embodiments provided in an embodiment of the present application. The communication device 1700 can specifically be a communication device as a network device in the above-mentioned embodiments. The example shown in Figure 17 is that the network device is implemented through the network device (or a component in the network device), wherein the structure of the communication device can refer to the structure shown in Figure 17.

[0388] The communication device 1700 includes at least one processor 1711 and at least one network interface 1714. Further optionally, the communication device also includes at least one memory 1717, at least one transceiver 1713 and one or more antennas 1715. The processor 1711, the memory 1717, the transceiver 1713 and the network interface 1714 are connected, for example, via a bus. In an embodiment of the present application, the connection may include various interfaces, transmission lines or buses, etc., which are not limited in this embodiment. The antenna 1715 is connected to the transceiver 1713. The network interface 1714 is used to enable the communication device to communicate with other communication devices through a communication link. For example, the network interface 1714 may include a network interface between the communication device and the core network device, such as an S1 interface, and the network interface may include a network interface between the communication device and other communication devices (such as other network devices or core network devices), such as an X2 or Xn interface.

[0389] Processor 1711 is primarily used to process communication protocols and communication data, control the entire communication device, execute software programs, and process software program data, 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. The baseband processor is primarily used to process communication protocols and communication data, while the central processing unit is primarily used to control the entire terminal device, execute software programs, and process software program data. Processor 1711 in Figure 17 may integrate the functions of both a baseband processor and a central processing unit. Those skilled in the art will appreciate that the baseband processor and the central processing unit may also be independent processors interconnected via a bus or other technology. Those skilled in the art will appreciate that a terminal device may include multiple baseband processors to accommodate different network standards, multiple central processing units to enhance its processing capabilities, and various components of the terminal device may be connected via various buses. The baseband processor may also be referred to as a baseband processing circuit or a baseband processing chip. The central processing unit may also be referred to as a central processing circuit or a central processing chip. The functionality for processing communication protocols and communication data may be built into the processor or stored in memory as a software program, which is executed by the processor to implement the baseband processing functionality.

[0390] The memory is mainly used to store software programs and data. The memory 1717 can exist independently and be connected to the processor 1711. Alternatively, the memory 1717 can be integrated with the processor 1711, for example, integrated into a single chip. The memory 1717 can store program codes for executing the technical solutions of the embodiments of the present application, and the execution is controlled by the processor 1711. The various computer program codes executed can also be regarded as drivers for the processor 1711.

[0391] Figure 17 shows only one memory and one processor. In an actual terminal device, there may be multiple processors and multiple memories. The memory may also be referred to as a storage medium or storage device. The memory may be a storage element on the same chip as the processor, i.e., an on-chip storage element, or an independent storage element, which is not limited in the present embodiment.

[0392] The transceiver 1713 can be used to support the reception or transmission of radio frequency signals between the communication device and the terminal. The transceiver 1713 can be connected to the antenna 1715. The transceiver 1713 includes a transmitter Tx and a receiver Rx. Specifically, one or more antennas 1715 can receive radio frequency signals. The receiver Rx of the transceiver 1713 is used to receive the radio frequency signal from the antenna, convert the radio frequency signal into a digital baseband signal or a digital intermediate frequency signal, and provide the digital baseband signal or digital intermediate frequency signal to the processor 1711 so that the processor 1711 can further process the digital baseband signal or digital intermediate frequency signal, such as demodulation and decoding. In addition, the transmitter Tx in the transceiver 1713 is also used to receive a modulated digital baseband signal or digital intermediate frequency signal from the processor 1711, convert the modulated digital baseband signal or digital intermediate frequency signal into a radio frequency signal, and transmit the radio frequency signal through one or more antennas 1715. Specifically, the receiver Rx can selectively perform one or more stages of down-mixing and analog-to-digital conversion on the RF signal to obtain a digital baseband signal or a digital intermediate frequency signal. The order of the down-mixing and analog-to-digital conversion processes is adjustable. The transmitter Tx can selectively perform one or more stages of up-mixing and digital-to-analog conversion on the modulated digital baseband signal or digital intermediate frequency signal to obtain a RF signal. The order of the up-mixing and digital-to-analog conversion processes is adjustable. The digital baseband signal and the digital intermediate frequency signal may be collectively referred to as digital signals.

[0393] The transceiver 1713 may also be referred to as a transceiver unit, a transceiver, a transceiver device, etc. Optionally, a device in the transceiver unit that implements a receiving function may be referred to as a receiving unit, and a device in the transceiver unit that implements a transmitting function may be referred to as a transmitting unit. That is, the transceiver unit includes a receiving unit and a transmitting unit. The receiving unit may also be referred to as a receiver, an input port, a receiving circuit, etc., and the transmitting unit may be referred to as a transmitter, a transmitter, or a transmitting circuit, etc.

[0394] It should be noted that the communication device 1700 shown in Figure 17 can be specifically used to implement the steps implemented by the network device in the aforementioned method embodiment, and to achieve the corresponding technical effects of the network device. The specific implementation method of the communication device 1700 shown in Figure 17 can refer to the description in the aforementioned method embodiment, and will not be repeated here.

[0395] An embodiment of the present application further provides a computer-readable storage medium storing one or more computer-executable instructions. When the computer-executable instructions are executed by a processor, the processor executes the method described in the possible implementation manner of the terminal device in the aforementioned embodiment.

[0396] An embodiment of the present application further provides a computer-readable storage medium storing one or more computer-executable instructions. When the computer-executable instructions are executed by a processor, the processor executes the method described in the possible implementation of the network device in the aforementioned embodiment.

[0397] An embodiment of the present application also provides a computer program product (or computer program) that stores one or more computers. When the computer program product is executed by the processor, the processor executes the method of the possible implementation method of the above-mentioned terminal device.

[0398] An embodiment of the present application also provides a computer program product storing one or more computers. When the computer program product is executed by the processor, the processor executes the method of the possible implementation of the above-mentioned network device.

[0399] An embodiment of the present application also provides a chip system, which includes at least one processor for supporting a communication device to implement the functions involved in the possible implementation methods of the above-mentioned communication device. Optionally, the chip system also includes an interface circuit, which provides program instructions and / or data to the at least one processor. In one possible design, the chip system may also include a memory for storing the necessary program instructions and data for the communication device. The chip system may be composed of a chip, or may include a chip and other discrete devices, wherein the communication device may specifically be a terminal device in the aforementioned method embodiment.

[0400] An embodiment of the present application also provides a chip system, which includes at least one processor for supporting a communication device to implement the functions involved in the possible implementation methods of the above-mentioned communication device. Optionally, the chip system also includes an interface circuit, which provides program instructions and / or data to the at least one processor. In one possible design, the chip system may also include a memory for storing the necessary program instructions and data for the communication device. The chip system may be composed of a chip, or may include a chip and other discrete devices, wherein the communication device may specifically be a network device in the aforementioned method embodiment.

[0401] An embodiment of the present application also provides a communication system, and the network system architecture includes the terminal device and network device in any of the above embodiments.

[0402] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0403] The units described as separate components may or may not be physically separate, and 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 these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0404] In addition, the functional units in the various embodiments of the present application can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit. If the integrated unit is implemented in the form of 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 the present application is essentially 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, which is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

Claims

1. A data processing method, characterized in that: include: The first device receives first data from the second device, where the first data is data processed by a first machine learning model; The first device processes the first data through a second machine learning model to obtain second data. The structure of the first machine learning model is the same as that of the second machine learning model. The first device and the second device are used to jointly perform data processing.

2. The method according to claim 1, characterized in that The second machine learning model is a diffusion model, and the second machine learning model is used to perform noise reduction processing on the first data.

3. The method according to claim 1 or 2, characterized in that The method further comprises: The first device receives first information from the second device, where the first information is used to request the first device to perform processing on the first data.

4. The method according to claim 3, characterized in that The first information is used to indicate that the number of times the first data is to be processed is the first number; The first device processes the first data using a second machine learning model to obtain second data, including: The first device processes the first data a first number of times through the second machine learning model to obtain the second data, wherein the capability of the first device supports completing the first number of processing of the first data.

5. The method according to claim 4, characterized in that The method further comprises: The first device sends the second data to the second device; Alternatively, the first device sends the second data to a source device, wherein the first information is further used to indicate information of the source device, and the source device is the first device to request assistance in processing data.

6. The method according to claim 3, characterized in that The first information is used to indicate that the number of times the first data is to be processed is the first number; The first device processes the first data using a second machine learning model to obtain second data, including: The first device processes the first data a second time using the second machine learning model to obtain the second data, wherein the first time is greater than the second time, and the capabilities of the first device do not support completing the first processing of the first data the first time; The method further comprises: The first device sends the second data and the second information to the third device; The second information is used to indicate that the number of times the second data is to be processed is the third number, the third number is the difference between the first number and the second number, and the third device is used to assist the first device in performing data processing.

7. The method according to claim 1 or 2, characterized in that The method further comprises: The first device sends assistance request information to the second device, where the assistance request information is used to request the second device to assist in processing data.

8. The method according to claim 1, 2 or 7, characterized in that The method further comprises: The first device sends third information to the central device, where the third information is used to indicate the number of times the first device needs to process the data; The first device receives feedback information from the central device, where the feedback information is used to indicate that the second device is an assisting node.

9. The method according to claim 8, characterized in that The method further comprises: The first device receives fourth information from the central device, where the fourth information indicates a number of times the first device needs to process the data received from the second device; The first device processes the first data using a second machine learning model to obtain second data, including: The first device processes the first data through the second machine learning model based on the fourth information to obtain the second data required by the first device.

10. The method according to claim 9, characterized in that The fourth information is further used to indicate information of a third device, where the third device is a device to receive the processed data; The method further comprises: The first device sends the second data to the third device according to the fourth information.

11. The method according to claim 7 or 8, characterized in that The method further comprises: The first device receives fifth information from the second device, where the fifth information indicates a number of times the first data has been processed; The first device processes the first data through the second machine learning model according to the number of processing times and the number of processing times of the data required by the first device to obtain the second data required by the first device.

12. A data processing method, characterized in that: include: The first device processes the original data using a first machine learning model to obtain first data; The first device sends the first data to the second device; The first device receives second data sent by a second device or other device, and the second data is obtained by processing the first data based on a second machine learning model, and the structure of the first machine learning model is the same as that of the second machine learning model.

13. The method according to claim 12, characterized in that The first machine learning model is a diffusion model, and the first machine learning model is used to perform noise reduction processing on the original data.

14. The method according to claim 12 or 13, characterized in that The method further comprises: The first device sends first information to the second device, where the first information is used to request the second device to perform processing on the first data, and / or the first information is used to indicate the number of times the first data is to be processed, where the number of times the first data is to be processed is determined based on the number of times the original data needs to be processed and the number of times the first device performs processing on the original data.

15. A data processing method, characterized in that: include: The central device receives first information from a first device and second information from a second device, wherein the first information indicates a first number of data processing times required by the first device, and the second information indicates a second number of data processing times required by the second device, wherein a data processing model corresponding to the first number of data processing times is the same as a data processing model corresponding to the second number of data processing times; The central device sends third information to the second device, where the third information is used to instruct the second device to send processed data to the first device, where the second processing times required for the data by the second device is less than or equal to the first processing times required for the data by the first device.

16. The method according to claim 15, characterized in that The method further comprises: The central device sends fourth information to the first device, where the fourth information is used to indicate the number of times the first device needs to process the data received from the second device.

17. A model training method, characterized in that: include: The first device obtains a training sample set, where the training sample set includes first data and second data, where the first data is obtained based on the second data, and the second data is a training label for the first data; The first device trains a first machine learning model based on the training sample set to obtain a trained first machine learning model, wherein the first machine learning model is used to process the first data; The first device sends the trained first machine learning model to the second device, and the second device is a device for aggregating machine learning models with the same structure and different parameters trained by multiple devices.

18. The method according to claim 17, characterized in that The method further comprises: The first device sends first information to a third device, where the first information is used to indicate capabilities related to model training on the first device, and the third device is used to determine training content that the multiple devices are responsible for based on the capabilities of the multiple devices participating in the machine learning model training; The first device receives second information from the third device, where the second information is used to indicate a number of times the first machine learning model trained on the first device processes input data, and the second information is further used to indicate input data requirements of the first machine learning model; The first device processes the original data according to the input data requirement indicated by the second information and the number of times the first machine learning model processes the input data to obtain the second data; The first device processes the second data according to the number of times the first machine learning model processes the input data indicated by the second information to obtain the first data.

19. The method according to claim 17 or 18, characterized in that The method further comprises: The first device receives a second machine learning model from the second device; The first device trains the second machine learning model based on the training sample set to obtain a trained second machine learning model; The first device sends the trained second machine learning model to the second device.

20. A model training method, characterized in that: include: The first device receives a plurality of capability information, the plurality of capability information coming from a plurality of different devices, and each capability information in the plurality of capability information is used to indicate a capability related to model training on the device; The first device sends different training configuration information to the multiple different devices according to the multiple capability information. The training configuration information is used to indicate the number of times the machine learning model trained on the device processes the input data. The training configuration information is also used to indicate the input data requirements of the machine learning model trained on the device. The machine learning models trained by the multiple different devices are models with the same structure.

21. A communication device, characterized in that: The communication device is a first device, and the communication device includes: a transceiver module, configured to receive first data from a second device, wherein the first data is data processed by a first machine learning model; A processing module is used to process the first data through a second machine learning model to obtain second data. The structure of the first machine learning model is the same as that of the second machine learning model. The first device and the second device are used to jointly perform data processing.

22. The device according to claim 21, characterized in that The second machine learning model is a diffusion model, and the second machine learning model is used to perform noise reduction processing on the first data.

23. The device according to claim 21 or 22, characterized in that The transceiver module is further configured to receive first information from the second device, where the first information is used to request the first device to process the first data.

24. The device according to claim 23, characterized in that The first information is used to indicate that the number of times the first data is to be processed is the first number; The processing module is also used to process the first data the first number of times through the second machine learning model to obtain the second data, wherein the capability of the first device supports completing the first number of processing of the first data.

25. The device according to claim 24, characterized in that Also includes: a sending module, configured to send the second data to the second device; Alternatively, the sending module is configured to send the second data to a source device, wherein the first information is further configured to indicate information of the source device, and the source device is the first device to request assistance in processing data.

26. The device according to claim 23, characterized in that The first information is used to indicate that the number of times the first data is to be processed is the first number; The processing module is further configured to process the first data a second time using the second machine learning model to obtain the second data, wherein the first time is greater than the second time, and the capability of the first device does not support completing the first processing of the first data the first time; The transceiver module is further configured to send the second data and the second information to a third device; The second information is used to indicate that the number of times the second data is to be processed is the third number, the third number is the difference between the first number and the second number, and the third device is used to assist the first device in performing data processing.

27. The device according to claim 21 or 22, characterized in that The transceiver module is further configured to send assistance request information to the second device, where the assistance request information is used to request the second device to assist in processing data.

28. The device according to claim 21, 22 or 27, characterized in that The transceiver module is further used for: sending third information to the central device, where the third information is used to indicate a number of data processing times required by the first device; Feedback information from the central device is received, where the feedback information is used to indicate that the second device is an assisting node.

29. The device according to claim 28, characterized in that The transceiver module is further configured to receive fourth information from the central device, wherein the fourth information is configured to indicate a number of times the first device needs to process the data received from the second device; The processing module is further used to process the first data through the second machine learning model according to the fourth information to obtain the second data required by the first device.

30. The device according to claim 29, characterized in that The fourth information is further used to indicate information of a third device, where the third device is a device to receive the processed data; The transceiver module is further configured to send the second data to the third device according to the fourth information.

31. The device according to claim 27 or 28, characterized in that The transceiver module is further configured to receive fifth information from the second device, where the fifth information indicates a number of times the first data has been processed; The processing module is further used to process the first data through the second machine learning model according to the number of processing times and the number of processing times of the data required by the first device to obtain the second data required by the first device.

32. The device according to any one of claims 21 to 31, characterized in that The transceiver module is a transceiver, and the processing module is a processor.

33. A communication device, characterized in that: The communication device is a first device, comprising: a processing module, configured to process the raw data using a first machine learning model to obtain first data; a transceiver module, configured to send the first data and the first information to the second device; The transceiver module is also used to receive second data sent by a second device or other device. The second data is obtained by processing the first data based on a second machine learning model, and the structure of the first machine learning model is the same as that of the second machine learning model.

34. The device according to claim 33, characterized in that The first machine learning model is a diffusion model, and the first machine learning model is used to perform noise reduction processing on the original data.

35. The device according to claim 33 or 34, characterized in that The transceiver module is also used to send first information to the second device, where the first information is used to request the second device to perform processing on the first data, and / or the first information is used to indicate the number of times the first data is to be processed, where the number of times the first data is to be processed is determined based on the number of times the original data needs to be processed and the number of times the first device performs processing on the original data.

36. The device according to any one of claims 33 to 35, characterized in that The transceiver module is a transceiver, and the processing module is a processor.

37. A communication device, characterized in that: include: a transceiver module configured to receive first information from a first device and second information from a second device, wherein the first information indicates a first number of data processing times required by the first device, and the second information indicates a second number of data processing times required by the second device, wherein a data processing model corresponding to the first number of data processing times is the same as a data processing model corresponding to the second number of data processing times; The transceiver module is further used to send third information to the second device, where the third information is used to instruct the second device to send processed data to the first device, wherein the second processing times of the data required by the second device is less than or equal to the first processing times of the data required by the first device.

38. The device according to claim 37, characterized in that The transceiver module is further configured to send fourth information to the first device, where the fourth information is configured to indicate a number of times the first device needs to process data received from the second device.

39. The device according to any one of claims 37-38, characterized in that The transceiver module is a transceiver.

40. A model training device, characterized in that include: a transceiver module, configured to obtain a training sample set, the training sample set including first data and second data, the first data being obtained based on the second data, and the second data being a training label for the first data; a processing module, configured to train a first machine learning model based on the training sample set to obtain a trained first machine learning model, wherein the first machine learning model is used to process the first data; The transceiver module is also used to send the trained first machine learning model to a second device, and the second device is a device for aggregating machine learning models with the same structure and different parameters trained by multiple devices.

41. The device according to claim 40, characterized in that The transceiver module is further configured to send first information to a third device, where the first information is used to indicate capabilities related to model training on the first device, and the third device is configured to determine training content for the multiple devices participating in the machine learning model training based on the capabilities of the multiple devices; The transceiver module is further configured to receive second information from the third device, where the second information is configured to indicate a number of times the first machine learning model trained on the first device processes input data, and the second information is further configured to indicate an input data requirement of the first machine learning model; The processing module is further configured to process the original data according to the input data requirement indicated by the second information and the number of times the first machine learning model processes the input data to obtain the second data; The processing module is further used to process the second data according to the number of times the first machine learning model processes the input data indicated by the second information to obtain the first data.

42. The device according to claim 40 or 41, characterized in that The transceiver module is further configured to receive a second machine learning model from the second device; The processing module is further configured to train the second machine learning model based on the training sample set to obtain a trained second machine learning model; The transceiver module is also used to send the trained second machine learning model to the second device.

43. The device according to any one of claims 40 to 42, characterized in that The transceiver module is a transceiver, and the processing module is a processor.

44. A model training device, characterized in that include: a transceiver module, configured to receive a plurality of capability information, the plurality of capability information being from a plurality of different devices, wherein each capability information in the plurality of capability information is used to indicate a capability on the device related to model training; The transceiver module is also used to send different training configuration information to the multiple different devices according to the multiple capability information. The training configuration information is used to indicate the number of times the machine learning model trained on the device processes the input data. The training configuration information is also used to indicate the input data requirements of the machine learning model trained on the device. The machine learning models trained by the multiple different devices are models with the same structure.

45. The device according to claim 44, characterized in that The transceiver module is a transceiver.

46. ​​A communication device, characterized in that comprising at least one processor coupled to a memory, The memory is used to store programs or instructions; The at least one processor is used to execute the program or instructions so that the device implements the method as described in any one of claims 1 to 11; or, implements the method as described in any one of claims 12 to 14; or, implements the method as described in any one of claims 15 to 16; or, implements the method as described in any one of claims 17 to 19; or, implements the method as described in claim 20.

47. A communication device, characterized in that comprising at least one processor; The at least one processor is used to execute programs or instructions so that the device implements the method as described in any one of claims 1 to 11; or, implements the method as described in any one of claims 12 to 14; or, implements the method as described in any one of claims 15 to 16; or, implements the method as described in any one of claims 17 to 19; or, implements the method as described in claim 20.

48. A communication system, characterized in that include: The communication device according to any one of claims 21 to 32 and the communication device according to any one of claims 33 to 36.

49. The system according to claim 48, wherein The system further comprises: a communication device according to any one of claims 37 to 39.

50. A communication system, characterized in that include: A model training device as described in any one of claims 40 to 43 and a model training device as described in any one of claims 44-45.

51. A computer-readable storage medium, characterized in that The readable storage medium stores instructions, which, when executed by a computer, cause the method described in any one of claims 1 to 11 to be executed; or, cause the method described in any one of claims 12 to 14 to be executed; or, cause the method described in any one of claims 15 to 16 to be executed; or, cause the method described in any one of claims 17 to 19 to be executed; or, cause the method described in claim 20 to be executed.

52. A computer program product, characterized in that The program product includes instructions, which, when executed on a computer, cause the method of any one of claims 1 to 11 to be executed; or, cause the method of any one of claims 12 to 14 to be executed; or, cause the method of any one of claims 15 to 16 to be executed; or, cause the method of any one of claims 17 to 19 to be executed; or, cause the method of claim 20 to be executed.

Images