Communication method and related apparatus

Abstract

Provided in the embodiments of the present application are a communication method and a related apparatus. The method comprises: a first apparatus measuring metrics of a first model on the basis of first data, so as to determine a first metric measurement result; the first apparatus filtering the first metric measurement result, so as to obtain a filtered metric measurement result; and the first apparatus sending the filtered metric measurement result to a second apparatus, or, the first apparatus managing the first model on the basis of the filtered metric measurement result, and sending a management result of the first model to the second apparatus. It can be seen therefrom that the first apparatus filters the first metric measurement result, so as to obtain the filtered metric measurement result, thereby facilitating an improvement to the accuracy of the metric measurement result, that is, the metrics of the first model are reflected more accurately. The first apparatus or the second apparatus efficiently manages the first model on the basis of the filtered metric measurement result, thereby avoiding unnecessary model management and thus improving the efficiency of model management.

Description

Communication method and related apparatus

[0001] This application claims priority from the Chinese patent application No. 202410722914.2 filed with the State Intellectual Property Office of China on June 5, 2024 and entitled "Communication method and related apparatus", the whole content of which is incorporated herein by reference. TECHNICAL FIELD

[0002] The present application relates to the technical field of wireless communication, in particular to a communication method and related apparatus. BACKGROUND

[0003] At present, artificial intelligence (AI) technology is introduced into a wireless communication system. The AI technology can be used for compression and reconstruction of wireless channel information, beam management, and positioning enhancement, etc. An AI model is obtained based on data training to improve the performance of completing a wireless task through the AI model. As shown in FIG. 1, a wireless AI framework includes data collection, model training, model management, model inference, and model storage. A device can measure an index of the AI model based on data to obtain the performance of the AI model. Optionally, the device performs management of the model according to the measured result. For example, switching of the AI model, or selection of the AI model, etc.

[0004] It can be seen that the index measurement of the AI model is related to data, and when the data is incorrect or is abnormal data, the index of the AI model measured by the device is inaccurate. Therefore, the device performs AI model management based on the measured index of the AI model, which can cause unnecessary AI model management, and the management efficiency of the AI model is low. For example, unnecessary AI model switching, retraining, or adjustment of model parameters, etc. SUMMARY

[0005] The present application provides a communication method and related apparatus, for a first device to measure an index of a first model according to first data, to determine a first index measurement result; and the first device filters the first index measurement result to obtain a filtered index measurement result. Thus, the inaccuracy of the measured index measurement result is avoided, or in other words, the accuracy of the measured index measurement result is improved, that is, the index of the first model is more accurately reflected. Then, the first device manages the first model according to the filtered index measurement result, thereby realizing efficient management of the first model. Alternatively, the first device sends the filtered index measurement result to a second device, so as to facilitate the second device to manage the first model efficiently in combination with the filtered index measurement result. This is beneficial to avoid unnecessary model management and improve the efficiency of model management.

[0006] The first aspect of the present application provides a communication method, which can be executed by a first device. The first device can be a terminal device, an access network device or a core network device, or a component (for example, a processor, a chip, or a chip system, etc.) in the terminal device, the access network device or the core network device, or a logic module or software capable of realizing all or part of the functions of the terminal device, or a logic module or software capable of realizing all or part of the functions of the access network device, or a logic module or software capable of realizing all or part of the functions of the core network device. The method comprises the following steps: the first device measures an index of a first model according to first data, to obtain a first index measurement result; the first device filters the first index measurement result to obtain a filtered index measurement result; the first device sends the filtered index measurement result to a second device, or the first device manages the first model according to the filtered index measurement result, and sends a management result of the first model to the second device.

[0007] According to the above technical solution, the first device filters the first index measurement result to obtain the filtered index measurement result. This is beneficial to avoid inaccurate index measurement results measured by the first device, or to avoid inaccurate index measurement results caused by noisy data or abnormal data. The filtered index measurement result can more accurately reflect the index of the first model. Then, the first device manages the first model according to the filtered index measurement result, thereby realizing efficient management of the first model. Alternatively, the first device sends the filtered index measurement result to the second device, which is beneficial to efficient management of the first model by the second device in combination with the filtered index measurement result. This is beneficial to avoid unnecessary model management and improve the efficiency of model management.

[0008] Based on the first aspect, in a possible implementation manner, the method further comprises: the first device receives first configuration information from the second device, the first configuration information being used to indicate a target filtering manner; and the first device filters the first index measurement result to obtain the filtered index measurement result, comprising: the first device filters the first index measurement result according to the target filtering manner to obtain the filtered index measurement result. The above implementation manner provides a scheme for filtering the first index measurement result by the first device, and realizes filtering of the first index measurement result by the first device based on the target filtering manner indicated by the first configuration information. This is beneficial to effective management of the first model by the second device in combination with the filtered index measurement result.

[0009] In a possible implementation manner of the first aspect, the target filtering manner is the first filtering manner, and the first filtering manner comprises weighting the first index measurement result and a historical filtered index measurement result, the historical filtered index measurement result comprising a filtered index measurement result obtained by filtering an index measurement result of the first model measured by the first device before the first index measurement result; or the target filtering manner is the second filtering manner, and the second filtering manner comprises weighting, by the first device, the first index measurement result and a historical index measurement result, the first index measurement result being obtained by measuring the index of the first model by the first device at the Nth time or the Nth time, the historical index measurement result comprising an index measurement result obtained by measuring the index of the first model by the first device at the K times before the Nth time, or an index measurement result obtained by measuring the index of the first model by the first device at the K times before the Nth time, K being an integer greater than or equal to 1, and N being an integer greater than or equal to 2.

[0010] In the implementation manners described above, the filtering is weighting, by the first device, the first index measurement result and the historical filtered index measurement result or the historical index measurement result to obtain the filtered index measurement result. This is advantageous to improve the accuracy of the filtered index measurement result, that is, to more accurately reflect the index of the first model. In the case that the first data is noisy data or abnormal data, the first index measurement result cannot accurately reflect the index of the first model, thereby avoiding unnecessary model management.

[0011] In a possible implementation manner of the first aspect, the first configuration information comprises a first weighting factor used for weighting the first index measurement result and the historical filtered index measurement result; or the first configuration information comprises at least one second weighting factor used for weighting the first index measurement result and the historical index measurement result, and / or K. The second device configures the corresponding weighting factor and the sliding window length K for the first device. Thus, the first device filters the first index measurement result. This avoids inaccurate index measurement results and affects the efficiency of model management.

[0012] In a possible implementation manner of the first aspect, the method further comprises: receiving, by the first device, second configuration information from the second device, the second configuration information being used for indicating a measurement configuration; and determining, by the first device, the first index measurement result according to the first data and the measurement configuration, comprising: measuring, by the first device, the index of the first model according to the first data and the measurement configuration, and determining the first index measurement result. In this implementation manner, the first device measures the index of the first model in combination with the measurement configuration, thereby obtaining the corresponding index measurement result. This facilitates subsequent filtering of the index measurement result and improves the accuracy of the index measurement result.

[0013] In a possible implementation of the first aspect, the measurement configuration comprises at least one of: an index to be measured, or a sample number used to determine the index to be measured. Thus, the first device can measure the corresponding index.

[0014] In a possible implementation of the first aspect, the index to be measured comprises at least one of: a model index, or a system performance, and the model index comprises at least one of: an accuracy of a model inference result, a loss function value of model training, a variation of a model parameter of model training, a validation data set performance, or an input-output distribution drift. This implementation shows some possible indexes to be measured, which facilitates implementation of the solution.

[0015] In a possible implementation of the first aspect, the first device measures the index of the first model according to the first data to determine the first index measurement result, comprising: the first device measures the index of the first model according to the first data to obtain a second index measurement result; the first device performs anomaly detection on the first data according to the second index measurement result to obtain abnormal data; and the first device determines the first index measurement result, which is the index measurement result excluding the index measurement result corresponding to the abnormal data in the second index measurement result. Thus, the index measurement result corresponding to the abnormal data is excluded, and the accuracy of the index measurement result is improved. This facilitates improvement of the efficiency of model management.

[0016] In a possible implementation of the first aspect, the first data comprises one or more data, and the abnormal data comprises data whose value is greater than or equal to a first threshold value from a reference data value. This provides a manner of determining abnormal data, which facilitates exclusion of the index measurement result corresponding to the abnormal data and improvement of the accuracy of the index measurement result. This facilitates improvement of the efficiency of model management.

[0017] In a possible implementation of the first aspect, the second index measurement result comprises one or more model indexes, and the index measurement result corresponding to the abnormal data is an index measurement result whose difference from a reference model index is greater than or equal to a second threshold value; or, the second index measurement result comprises one or more system performances, and the index measurement result corresponding to the abnormal data is an index measurement result whose difference from a reference system performance is greater than or equal to a third threshold value. This implementation shows two possible implementations of the index measurement result corresponding to the abnormal data, which facilitates implementation of the solution. Thus, the index measurement result corresponding to the abnormal data is excluded, and the accuracy of the index measurement result is improved. This facilitates improvement of the efficiency of model management.

[0018] In a possible implementation manner of the first aspect, the method further includes: the first device sending an abnormal result to the second device, the abnormal result including abnormal data and / or a proportion of the abnormal data. This facilitates the second device to obtain the abnormal result and manage the first model in combination with the abnormal result. This improves the efficiency of model management.

[0019] In a possible implementation manner of the first aspect, the first device manages the first model according to the filtered index measurement result, including: loading a second model, the first model being obtained by training the AI service based on the first data for the Mth time, the second model being a model obtained by training the AI service for the (M-1)th time, M being an integer greater than or equal to 2; or updating a model parameter of the first model. In this implementation manner, some possible management operations of the first device on the first model in the case of abnormal data are shown. For example, if abnormal data occurs, the second model obtained by the previous training can be loaded. For another example, if the number of abnormal times is large, the model parameter of the first model can be adjusted. This facilitates improving the performance of model training or the performance of model inference.

[0020] In a possible implementation manner of the first aspect, the method further includes: the first device receiving third configuration information from the second device, the third configuration information being used to indicate reporting configuration; and the first device sending the filtered index measurement result to the second device, including: the first device sending the filtered index measurement result to the second device according to the reporting configuration. This facilitates the first device to report appropriate index measurement results, and facilitates the second device to manage the first model in combination with the filtered index measurement result. This improves the efficiency of model management.

[0021] In a possible implementation manner of the first aspect, the reporting configuration includes at least one of the following: reporting condition or reporting period.

[0022] In a possible implementation manner of the first aspect, the first device manages the first model according to the filtered index measurement result, including: the first device performing inference management or training management on the first model according to the filtered index measurement result. For the first model, the first device can perform inference management or training management. This implements efficient management of the first model. This avoids unnecessary model management operations.

[0023] In a possible implementation manner of the first aspect, the filtered indicator measurement result includes a model indicator and / or system performance of the first model; the inference management includes: if the model indicator and / or system performance is less than or equal to a fourth threshold value, falling back to a default mode, the default mode being using a non-AI mode or a default AI model for transmission, or measuring an indicator of a third model, or switching to a fourth model; or, if the model indicator and / or system performance is greater than or equal to a fifth threshold value, activating the first model or switching to the first model; or, if the model indicator and / or system performance of the first model measured by the first device for Q consecutive times is less than or equal to a sixth threshold value, requesting to train the first model, or requesting to collect a data set, the data set being used for training the first model, Q being an integer greater than or equal to 1. In this implementation manner, the inference management is performed in combination with the actually measured indicator of the first model. This is beneficial to performing appropriate inference management operation on the first model, and improves the efficiency of model management.

[0024] In a possible implementation manner of the first aspect, the filtered indicator measurement result includes a model indicator and / or system performance of the first model; the training management includes: if the model indicator and / or system performance is less than or equal to a seventh threshold value, terminating the training of the first model, or adjusting a learning rate of the first model; or, if the model indicator and / or system performance is greater than or equal to an eighth threshold value, initializing the first model or loading a second model, the first model being obtained by performing M rounds of training on the first data for the AI service, the second model being a model obtained by performing M-1 rounds of training for the AI service, M being an integer greater than or equal to 2. In this implementation manner, the training management is performed in combination with the actually measured indicator of the first model. This is beneficial to performing appropriate training management operation on the first model, and improves the efficiency of model management.

[0025] The second aspect of the present application provides a communication method, which can be performed by a second device. The second device can be a terminal device, an access network device or a core network device, or a component (for example, a processor, a chip or a chip system) in the terminal device, the access network device or the core network device, or a logic module or software capable of realizing all or part of the functions of the terminal device, the access network device or the core network device. The method includes: receiving, by the second device, a filtered indicator measurement result from a first device, the filtered indicator measurement result being obtained by filtering a first indicator measurement result, the first indicator measurement result being determined according to a first data measurement result of a first model; and managing, by the second device, the first model according to the filtered indicator measurement result.

[0026] According to the above technical solution, the first index measurement result is determined according to the first data measurement of the index of the first model, and the filtered index measurement result is obtained by filtering the first index measurement result. The filtered index measurement result can more accurately reflect the index of the first model. The second device manages the first model according to the filtered index measurement result, and efficient management of the first model can be achieved. This is beneficial to avoid unnecessary model management and improve the efficiency of model management.

[0027] The third aspect of the present application provides a communication method, which can be executed by a second device. The second device can be a terminal device, an access network device, or a core network device, or a component (for example, a processor, a chip, or a chip system, etc.) in the terminal device, the access network device, or the core network device, or a logic module or software capable of realizing all or part of the functions of the terminal device, or a logic module or software capable of realizing all or part of the functions of the access network device, or a logic module or software capable of realizing all or part of the functions of the core network device. The method comprises: the second device receives a management result of a first model from a first device, the management result being obtained by managing the first model according to a filtered index measurement result, the filtered index measurement result being obtained by filtering a first index measurement result, and the first index measurement result being determined according to first data measurement of an index of the first model.

[0028] According to the above technical solution, the first index measurement result is determined according to the first data measurement of the index of the first model, and the filtered index measurement result is obtained by filtering the first index measurement result. The filtered index measurement result can more accurately reflect the index of the first model. The second device receives a management result of a first model from a first device, the management result being obtained by managing the first model according to a filtered index measurement result. This realizes efficient management of the first model by the first device. This is beneficial to avoid unnecessary model management and improve the efficiency of model management.

[0029] Based on the third aspect, in a possible implementation manner, the method further comprises: the second device sends a management confirmation to the first device, the management confirmation being used to determine the management result. This facilitates the first device to perform corresponding management on the first model. This avoids the problem of wireless transmission between the first device and the second device based on the first model in the case that the first device does not agree with the management operation of the first device.

[0030] In a possible implementation manner of the second aspect or the third aspect, the method further includes: sending, by the second device, first configuration information to the first device, the first configuration information being used to indicate a target filtering manner, and the target filtering manner being used to filter the first index measurement result. Thus, the first device filters the first index measurement result based on the target filtering manner. In this way, the inaccuracy of the first index measurement result caused by the first data being noisy data or abnormal data is avoided, and the efficiency of model management is affected.

[0031] In a possible implementation manner of the second aspect or the third aspect, the target filtering manner is a first filtering manner, and the first filtering manner includes performing weighted processing on the first index measurement result and a historical filtered index measurement result, the historical filtered index measurement result including a filtered index measurement result obtained by filtering an index measurement result of the first model measured by the first device before the first index measurement result; or the target filtering manner is a second filtering manner, and the second filtering manner includes performing weighted processing on the first index measurement result and a historical index measurement result, the first index measurement result being obtained by measuring the index of the first model by the first device at the Nth time or the Nth time, the historical index measurement result including an index measurement result obtained by measuring the index of the first model by the first device at the Kth time before the Nth time, or including an index measurement result obtained by measuring the index of the first model by the first device at the Kth time before the Nth time, K being an integer greater than or equal to 1, and N being an integer greater than or equal to 2. In the implementation manner, the filtering is to perform weighted processing on the first index measurement result and the historical filtered index measurement result or the historical index measurement result to obtain a filtered index measurement result. This is beneficial to improving the accuracy of the filtered index measurement result, that is, more accurately reflecting the index of the first model. In this way, unnecessary model management caused by the first index measurement result failing to accurately reflect the index of the first model in the case that the first data is noisy data or abnormal data is avoided, and the efficiency of model management is affected.

[0032] In a possible implementation manner of the second aspect or the third aspect, the method further includes: sending, by the second device, second configuration information to the first device, the second configuration information being used to indicate a measurement configuration, and the measurement configuration being used to measure the index of the first model by the first device. Thus, the first device measures the index of the first model based on the measurement configuration, and thus obtains a corresponding index measurement result.

[0033] In a possible implementation manner of the second aspect or the third aspect, the measurement configuration includes at least one of the following: an index to be measured, or a sample number used to determine the index to be measured. Thus, the first device measures a corresponding index.

[0034] In a possible implementation of the second aspect or the third aspect, the indicators to be measured include at least one of the following: a model indicator, or a system performance, and the model indicator includes at least one of the following: an accuracy of a model inference result, a loss function value of model training, a variation of a model parameter of model training, a validation data set performance, or an input-output distribution drift. This implementation shows some possible indicators to be measured, which facilitates implementation of the solutions.

[0035] In a possible implementation of the second aspect or the third aspect, the method further includes: receiving, by the second device, an abnormal result from the first device, the abnormal result including abnormal data, and / or a proportion of the abnormal data, and the abnormal data being abnormal data in the first data. This facilitates the second device to manage the first model in combination with the abnormal result. This improves efficiency of model management.

[0036] In a possible implementation of the second aspect or the third aspect, the second device manages the first model according to the filtered indicator measurement result, including: loading a second model, the first model being obtained by performing M rounds of training on the first data for the AI service, the second model being a model obtained by performing M-1 rounds of training for the AI service, and M being an integer greater than or equal to 2; or updating a model parameter of the first model. This implementation shows some possible management operations of the second device on the first model in the case of abnormal data. For example, if abnormal data occurs, the second model obtained by the previous round of training can be loaded. For another example, if the number of abnormal data is large, the model parameter of the first model can be adjusted. This facilitates improvement of performance of model training or performance of model inference.

[0037] In a possible implementation of the second aspect or the third aspect, the method further includes: sending, by the second device, third configuration information to the first device, the third configuration information being used to indicate reporting configuration, and the reporting configuration being used to report the filtered indicator measurement result. This facilitates the first device to report appropriate indicator measurement results, and facilitates the second device to manage the first model in combination with the filtered indicator measurement result. This improves efficiency of model management.

[0038] In a possible implementation of the second aspect or the third aspect, the second device manages the first model according to the filtered indicator measurement result, including: performing, by the second device, inference management or training management on the first model according to the filtered indicator measurement result. For the first model, the second device can perform inference management or training management. This implements efficient management of the first model. This avoids unnecessary model management operations.

[0039] In a possible implementation manner of the second aspect or the third aspect, the filtered indicator measurement result includes a model indicator and / or system performance of the first model; the inference management includes: if the model indicator and / or system performance is less than or equal to a fourth threshold value, falling back to a default mode, the default mode being using a non-AI mode or a default AI model for transmission, or measuring an indicator of a third model, or switching to a fourth model; or, if the model indicator and / or system performance is greater than or equal to a fifth threshold value, activating the first model or switching to the first model; or, if the model indicator and / or system performance of the first model measured by the first device for Q consecutive times is less than or equal to a sixth threshold value, requesting to train the first model or requesting to collect a data set, the data set being used for training the first model, Q being an integer greater than or equal to 1. In this implementation manner, the corresponding inference management is performed in combination with the actually measured indicator of the first model. This is beneficial to performing appropriate inference management operation on the first model, and improves the efficiency of model management.

[0040] In a possible implementation manner of the second aspect or the third aspect, the filtered indicator measurement result includes a model indicator and / or system performance of the first model; the training management includes: if the model indicator and / or system performance is less than or equal to a seventh threshold value, terminating the training of the first model or adjusting a learning rate of the first model; or, if the model indicator and / or system performance is greater than or equal to an eighth threshold value, initializing the first model or loading a second model, the first model being obtained by performing M rounds of training on the first data for the AI service, the second model being a model obtained by performing M-1 rounds of training for the AI service, M being an integer greater than or equal to 2. In this implementation manner, the corresponding training management is performed in combination with the actually measured indicator of the first model. This is beneficial to performing appropriate training management operation on the first model, and improves the efficiency of model management.

[0041] The fourth aspect of the present application provides a first device, including:

[0042] a processing module, configured to measure an indicator of a first model according to first data, determine a first indicator measurement result, and filter the first indicator measurement result to obtain a filtered indicator measurement result; a transceiver, configured to send the filtered indicator measurement result to a second device; or the processing module is further configured to manage the first model according to the filtered indicator measurement result; and the transceiver is further configured to send a management result of the first model to the second device.

[0043] In a possible implementation manner of the fourth aspect, the transceiver is further configured to receive first configuration information from the second device, the first configuration information being used to indicate a target filtering manner; and the processing module is specifically configured to filter the first indicator measurement result according to the target filtering manner to obtain the filtered indicator measurement result.

[0044] In a possible implementation manner based on the fourth aspect, the target filtering manner is the first filtering manner, and the first filtering manner comprises weighting the first index measurement result and a historical filtered index measurement result, the historical filtered index measurement result comprising a filtered index measurement result obtained by filtering an index measurement result of the first model measured by the first device before the first index measurement result; or the target filtering manner is the second filtering manner, and the second filtering manner comprises weighting, by the first device, the first index measurement result and a historical index measurement result, the first index measurement result being obtained by measuring the index of the first model by the first device at the Nth time or the Nth time, the historical index measurement result comprising an index measurement result obtained by measuring the index of the first model by the first device at the K previous times of the Nth time, or an index measurement result obtained by measuring the index of the first model by the first device at the K previous times of the Nth time, K being an integer greater than or equal to 1, and N being an integer greater than or equal to 2.

[0045] In a possible implementation manner based on the fourth aspect, the first configuration information comprises a first weighting factor used for weighting the first index measurement result and the historical filtered index measurement result; or the first configuration information comprises at least one second weighting factor used for weighting the first index measurement result and the historical index measurement result, and / or K.

[0046] In a possible implementation manner based on the fourth aspect, the transceiver is further configured to receive second configuration information from the second device, the second configuration information being used for indicating a measurement configuration; and the processor is specifically configured to measure the index of the first model according to the first data and the measurement configuration, and determine the first index measurement result.

[0047] In a possible implementation manner based on the fourth aspect, the measurement configuration comprises at least one of the following: an index to be measured, or a sample number used for determining the index to be measured.

[0048] In a possible implementation manner based on the fourth aspect, the index to be measured comprises at least one of the following: a model index, or a system performance, the model index comprising at least one of the following: an accuracy of a model inference result, a loss function value of model training, a variation of a model parameter of model training, a validation data set performance, or an input-output distribution drift.

[0049] In a possible implementation manner based on the fourth aspect, the processor is specifically configured to measure the index of the first model according to the first data, to obtain a second index measurement result; perform anomaly detection on the first data according to the second index measurement result, to obtain abnormal data; and determine the first index measurement result, the first index measurement result being an index measurement result other than an index measurement result corresponding to the abnormal data in the second index measurement result.

[0050] In a possible implementation manner based on the fourth aspect, the first data includes one or more data, and the abnormal data includes data whose value in the one or more data is greater than or equal to a first threshold value.

[0051] In a possible implementation manner based on the fourth aspect, the second index measurement result includes one or more model indexes, and the index measurement result corresponding to the abnormal data is an index measurement result whose difference from a reference model index is greater than or equal to a second threshold value; or, the second index measurement result includes one or more system performances, and the index measurement result corresponding to the abnormal data is an index measurement result whose difference from a reference system performance is greater than or equal to a third threshold value.

[0052] In a possible implementation manner based on the fourth aspect, the transceiver is further configured to: send, to the second device, the abnormal result, the abnormal result including the abnormal data, and / or a proportion of the abnormal data.

[0053] In a possible implementation manner based on the fourth aspect, the processing module is specifically configured to: load a second model, the first model being obtained by training the AI service based on the first data for the Mth time, the second model being a model obtained by training the AI service for the (M-1)th time, and M being an integer greater than or equal to 2; or update a model parameter of the first model.

[0054] In a possible implementation manner based on the fourth aspect, the transceiver is further configured to: receive third configuration information from the second device, the third configuration information being used to indicate a reporting configuration; and send, to the second device, the filtered index measurement result according to the reporting configuration.

[0055] In a possible implementation manner based on the fourth aspect, the reporting configuration includes at least one of the following: a reporting condition or a reporting period.

[0056] In a possible implementation manner based on the fourth aspect, the processing module is specifically configured to: perform inference management or training management on the first model according to the filtered index measurement result.

[0057] In a possible implementation manner based on the fourth aspect, the filtered index measurement result includes a model index and / or system performance of the first model; the inference management includes: if the model index and / or system performance is less than or equal to a fourth threshold value, falling back to a default mode, the default mode being to perform transmission in a non-AI manner, or measuring an index of a third model, or switching to a fourth model; or, if the model index and / or system performance is greater than or equal to a fifth threshold value, activating the first model, or switching to the first model; or, if the model index and / or system performance of the first model measured by the first device for Q consecutive times is less than or equal to a sixth threshold value, requesting to train the first model, or requesting to collect a data set, the data set being used for training the first model, Q being an integer greater than or equal to 1.

[0058] In a possible implementation manner based on the fourth aspect, the filtered index measurement result includes a model index and / or system performance of the first model; the training management includes: if the model index and / or system performance is less than or equal to a seventh threshold value, terminating the training of the first model, or adjusting a learning rate of the first model; or, if the model index and / or system performance is greater than or equal to an eighth threshold value, initializing the first model or loading a second model, the first model being obtained by performing M rounds of training based on the first data for the AI service, the second model being a model obtained by performing M-1 rounds of training for the AI service, M being an integer greater than or equal to 2.

[0059] The fifth aspect of the present application provides a second device, including:

[0060] The transceiver module is configured to receive a filtered index measurement result from the first device, the filtered index measurement result being obtained by filtering a first index measurement result, the first index measurement result being determined according to measurement of an index of the first model based on the first data;

[0061] The processing module is configured to manage the first model according to the filtered index measurement result.

[0062] The sixth aspect of the present application provides a second device, including:

[0063] The transceiver module is configured to receive a management result of the first model from the first device, the management result being obtained by managing the first model according to a filtered index measurement result, the filtered index measurement result being obtained by filtering a first index measurement result, the first index measurement result being determined according to measurement of an index of the first model based on the first data.

[0064] In a possible implementation manner based on the sixth aspect, the transceiver module is further configured to send a management confirmation to the first device, the management confirmation being used to determine the management result.

[0065] In a possible implementation manner based on the fifth aspect or the sixth aspect, the transceiver is further configured to: send, to the first device, first configuration information, the first configuration information being used to indicate a target filtering manner, the target filtering manner being used to filter the first index measurement result.

[0066] In a possible implementation manner based on the fifth aspect or the sixth aspect, the target filtering manner is a first filtering manner, the first filtering manner comprising performing weighted processing on the first index measurement result and a historical filtered index measurement result, the historical filtered index measurement result comprising a filtered index measurement result obtained by filtering an index measurement result of the first model measured by the first device before the first index measurement result; or the target filtering manner is a second filtering manner, the second filtering manner comprising performing weighted processing on the first index measurement result and a historical index measurement result, the first index measurement result being obtained by measuring the index of the first model by the first device at the Nth time or the Nth time, the historical index measurement result comprising an index measurement result obtained by measuring the index of the first model by the first device at the K times before the Nth time, or comprising an index measurement result obtained by measuring the index of the first model by the first device at the K times before the Nth time, K being an integer greater than or equal to 1, and N being an integer greater than or equal to 2.

[0067] In a possible implementation manner based on the fifth aspect or the sixth aspect, the transceiver is further configured to: send, to the first device, second configuration information, the second configuration information being used to indicate a measurement configuration, the measurement configuration being used for the first device to measure the index of the first model.

[0068] In a possible implementation manner based on the fifth aspect or the sixth aspect, the measurement configuration comprises at least one of the following: an index to be measured, or a sample number used to determine the index to be measured.

[0069] In a possible implementation manner based on the fifth aspect or the sixth aspect, the index to be measured comprises at least one of the following: a model index, or a system performance, the model index comprising at least one of the following: an accuracy of a model inference result, a loss function value of model training, a variation of a model parameter of model training, a validation data set performance, or an input-output distribution drift.

[0070] In a possible implementation manner based on the fifth aspect or the sixth aspect, the transceiver is further configured to: receive, from the first device, an abnormal result, the abnormal result comprising abnormal data and / or a proportion of the abnormal data, the abnormal data being abnormal data in the first data.

[0071] In a possible implementation manner of the fifth aspect or the sixth aspect, the processing module is specifically configured to: load the second model, the first model being obtained by training the AI service based on the first data for M rounds, the second model being a model obtained by training the AI service for M-1 rounds, M being an integer greater than or equal to 2; or update the model parameters of the first model.

[0072] In a possible implementation manner of the fifth aspect or the sixth aspect, the transceiving module is further configured to: send, to the first device, third configuration information, the third configuration information being used to indicate reporting configuration, the reporting configuration being used to report the filtered index measurement result.

[0073] In a possible implementation manner of the fifth aspect or the sixth aspect, the processing module is further configured to: perform inference management or training management on the first model according to the filtered index measurement result.

[0074] In a possible implementation manner of the fifth aspect or the sixth aspect, the filtered index measurement result includes a model index and / or system performance of the first model; the inference management includes: if the model index and / or system performance is less than or equal to a fourth threshold value, falling back to a default mode, the default mode being to perform transmission in a non-AI manner or a default AI model, or measuring an index of a third model, or switching to a fourth model; or if the model index and / or system performance is greater than or equal to a fifth threshold value, activating the first model or switching to the first model; or if the first device continuously measures the index and / or system performance of the first model Q times and the index and / or system performance is less than or equal to a sixth threshold value, requesting to train the first model or requesting to collect a data set, the data set being used to train the first model, Q being an integer greater than or equal to 1.

[0075] In a possible implementation manner of the fifth aspect or the sixth aspect, the filtered index measurement result includes a model index and / or system performance of the first model; the training management includes: if the model index and / or system performance is less than or equal to a seventh threshold value, terminating the training of the first model or adjusting a learning rate of the first model; or if the model index and / or system performance is greater than or equal to an eighth threshold value, initializing the first model or loading a second model, the first model being obtained by training the AI service based on the first data for M rounds, the second model being a model obtained by training the AI service for M-1 rounds, M being an integer greater than or equal to 2.

[0076] For the fourth aspect, the first apparatus can be a terminal device, an access network device, or a core network device, or a component (e.g., a processor, a chip, or a chip system, etc.) of the terminal device, the access network device, or the core network device, or a logic module or software capable of realizing all or part of the terminal device function, or a logic module or software capable of realizing all or part of the access network device function, or a logic module or software capable of realizing all or part of the core network device function. The transceiver module can be a transceiver, or an input / output interface; and the processing module can be a processor.

[0077] In an implementation manner, the first apparatus is a chip, a chip system, or a circuit configured in the terminal device. When the first apparatus is a chip, a chip system, or a circuit configured in the terminal device, the transceiver module can be an input / output interface, an interface circuit, an output circuit, an input circuit, a pin, or a related circuit on the chip, the chip system, or the circuit; and the processing module can be a processor, a processing circuit, or a logic circuit.

[0078] In an implementation manner, the first apparatus is a chip, a chip system, or a circuit configured in the access network device. When the first apparatus is a chip, a chip system, or a circuit configured in the access network device, the transceiver module can be an input / output interface, an interface circuit, an output circuit, an input circuit, a pin, or a related circuit on the chip, the chip system, or the circuit; and the processing module can be a processor, a processing circuit, or a logic circuit.

[0079] In an implementation manner, the first apparatus is a chip, a chip system, or a circuit configured in the core network device. When the first apparatus is a chip, a chip system, or a circuit configured in the core network device, the transceiver module can be an input / output interface, an interface circuit, an output circuit, an input circuit, a pin, or a related circuit on the chip, the chip system, or the circuit; and the processing module can be a processor, a processing circuit, or a logic circuit.

[0080] For the fifth aspect or the sixth aspect, the second apparatus can be a terminal device, an access network device, or a core network device, or a component (e.g., a processor, a chip, or a chip system, etc.) of the terminal device, the access network device, or the core network device, or a logic module or software capable of realizing all or part of the terminal device function, or a logic module or software capable of realizing all or part of the access network device function, or a logic module or software capable of realizing all or part of the core network device function. The transceiver module can be a transceiver, or an input / output interface; and the processing module can be a processor.

[0081] In an implementation, the second device is a chip, chip system or circuit configured in the terminal device. When the second device is a chip, chip system or circuit configured in the terminal device, the transceiving module can be an input / output interface, interface circuit, output circuit, input circuit, pin or related circuit on the chip, chip system or circuit; and the processing module can be a processor, processing circuit or logic circuit.

[0082] In an implementation, the second device is a chip, chip system or circuit configured in the access network device. When the second device is a chip, chip system or circuit configured in the access network device, the transceiving module can be an input / output interface, interface circuit, output circuit, input circuit, pin or related circuit on the chip, chip system or circuit; and the processing module can be a processor, processing circuit or logic circuit.

[0083] In an implementation, the second device is a chip, chip system or circuit configured in the core network device. When the second device is a chip, chip system or circuit configured in the core network device, the transceiving module can be an input / output interface, interface circuit, output circuit, input circuit, pin or related circuit on the chip, chip system or circuit; and the processing module can be a processor, processing circuit or logic circuit.

[0084] The seventh aspect of the present application provides a first device, which comprises a processor and a memory. The memory stores a computer program or computer instruction, and the processor is configured to invoke and run the computer program or computer instruction stored in the memory, so that the processor implements any one of the implementation manners of the first aspect.

[0085] Optionally, the first device further comprises a transceiver, and the processor is configured to control the transceiver to transceive signals.

[0086] The eighth aspect of the present application provides a second device, which comprises a processor and a memory. The memory stores a computer program or computer instruction, and the processor is configured to invoke and run the computer program or computer instruction stored in the memory, so that the processor implements any one of the implementation manners of the second aspect.

[0087] Optionally, the second device further comprises a transceiver, and the processor is configured to control the transceiver to transceive signals.

[0088] The ninth aspect of the present application provides a first device, which comprises a processor and an interface circuit, and the processor is configured to communicate with other devices through the interface circuit and perform the method described in the first aspect. The processor comprises one or more.

[0089] The tenth aspect of the present application provides a second device, comprising a processor and an interface circuit, the processor being configured to communicate with other devices through the interface circuit and perform the method described in the second aspect or the third aspect. The processor comprises one or more.

[0090] The eleventh aspect of the present application provides a first device, comprising a processor configured to be connected with a memory and invoke a program stored in the memory to perform the method described in the first aspect. The memory can be located in the first device or outside the first device. The processor comprises one or more.

[0091] The twelfth aspect of the present application provides a second device, comprising a processor configured to be connected with a memory and invoke a program stored in the memory to perform the method described in the second aspect or the third aspect. The memory can be located in the second device or outside the second device. The processor comprises one or more.

[0092] In an implementation manner, the first device in the first aspect, the fourth aspect can be a chip or a chip system. The second device in the second aspect, the third aspect, the fifth aspect, the sixth aspect can be a chip or a chip system.

[0093] The thirteenth aspect of the present application provides a computer program product comprising computer instructions, which, when executed on a computer, cause the computer to perform any implementation manner of any one of the first aspect to the third aspect.

[0094] The fourteenth aspect of the present application provides a computer readable storage medium comprising computer instructions, which, when executed on a computer, cause the computer to perform any implementation manner of any one of the first aspect to the third aspect.

[0095] The fifteenth aspect of the present application provides a chip device comprising a processor configured to invoke a computer program or computer instructions in a memory to cause the processor to perform any implementation manner of any one of the first aspect to the third aspect.

[0096] Optionally, the processor is coupled with the memory through an interface.

[0097] The sixteenth aspect of the present application provides a communication system, comprising the first device in the first aspect and the second device in the second aspect; or the first device in the first aspect and the second device in the third aspect.

[0098] As described in the above technical solution, the first device measures the indicators of the first model based on the first data and determines the measurement result of the first indicator. Then, the first device filters the measurement result of the first indicator to obtain a filtered measurement result. The first device sends the filtered measurement result to the second device, or the first device manages the first model based on the filtered measurement result and sends the management result of the first model to the second device. Therefore, the first device filters the measurement result of the first indicator to obtain a filtered measurement result. This avoids inaccurate measurement results, or rather, improves the accuracy of the measured indicator results, that is, more accurately reflects the indicators of the first model. Then, the first device manages the first model based on the filtered measurement result, achieving efficient management of the first model. Alternatively, the first device sends the filtered measurement result to the second device, facilitating the second device's efficient management of the first model in conjunction with the filtered measurement result. This helps avoid unnecessary model management and improves the efficiency of model management. Attached Figure Description

[0099] Figure 1 is a schematic diagram of a wireless AI framework according to an embodiment of this application;

[0100] Figure 2A is a schematic diagram of a communication system according to an embodiment of this application;

[0101] Figure 2B is another schematic diagram of the communication system according to an embodiment of this application;

[0102] Figure 3 is a schematic diagram of an embodiment of the communication method of this application;

[0103] Figure 4 is a flowchart illustrating a communication method according to an embodiment of this application;

[0104] Figure 5 is a schematic diagram of a first filtering method according to an embodiment of this application;

[0105] Figure 6 is a schematic diagram of a second filtering method according to an embodiment of this application;

[0106] Figure 7 is another flowchart illustrating the communication method according to an embodiment of this application;

[0107] Figure 8 is a structural schematic diagram of the first device according to an embodiment of this application;

[0108] Figure 9 is a structural schematic diagram of the second device according to an embodiment of this application;

[0109] Figure 10 is another structural schematic diagram of the second device according to an embodiment of this application;

[0110] Figure 11 is a schematic diagram of the structure of a device according to an embodiment of this application;

[0111] Figure 12 is a structural schematic diagram of a terminal device according to an embodiment of this application;

[0112] Figure 13 is a structural schematic diagram of a network device according to an embodiment of the present application. DETAILED DESCRIPTION

[0113] The embodiment of the present application provides a communication method and related apparatus, for a first device to measure an index of a first model according to first data, to determine a first index measurement result; the first device filters the first index measurement result to obtain a filtered index measurement result. Thus, the accuracy of the measured index measurement result is improved, that is, the index of the first model is more accurately reflected. Then, the first device manages the first model according to the filtered index measurement result, to realize efficient management of the first model. Alternatively, the first device sends the filtered index measurement result to a second device, to facilitate the second device to manage the first model efficiently in combination with the filtered index measurement result. This is beneficial to avoid unnecessary model management, and to improve the efficiency of model management.

[0114] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person skilled in the art without any creative work fall within the protection scope of the present application.

[0115] In the present application, the reference to “one embodiment” or “some embodiments” and the like means that the specific features, structures or characteristics described in connection with the embodiment are included in one or more embodiments of the present application. Thus, the statements “in one embodiment”, “in some embodiments”, “in other some embodiments”, “in further some embodiments” and the like appearing in different places in the specification are not necessarily all referring to the same embodiment, but mean “one or more but not all embodiments”, unless otherwise specifically emphasized. The terms “include”, “contain”, “have” and their variants mean “including but not limited to”, unless otherwise specifically emphasized.

[0116] In the description of the present application, unless otherwise specified, " / " means "or", for example, A / B can mean A or B. "And / or" herein is only a description of the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B can mean: A exists alone, A and B exist together, and B exists alone. In addition, "at least one" means one or more, and "multiple" means two or more. "At least one of the following" or the like means any combination of the items, including any combination of single item or multiple items. For example, at least one of a, b, or c can mean: a, b, c; a and b; a and c; b and c; or a and b and c. Where a, b, and c can be single or multiple.

[0117] It can be understood that in the present application, "indication" can include direct indication, indirect indication, display indication, and implicit indication. When describing that certain indication information is used to indicate A, it can be understood that the indication information carries A, directly indicates A, or indirectly indicates A.

[0118] In the present application, the information indicated by the indication information is referred to as the to-be-indicated information. In the specific implementation process, there are many ways to indicate the to-be-indicated information, for example, but not limited to, the to-be-indicated information can be directly indicated, such as the to-be-indicated information itself or an index of the to-be-indicated information, or the to-be-indicated information can be indirectly indicated by indicating other information, where the other information and the to-be-indicated information have an association relationship. It can also only indicate part of the to-be-indicated information, and the other part of the to-be-indicated information is known or agreed in advance. For example, the indication of a specific information can also be achieved by means of the pre-agreed (for example, the protocol stipulates) arrangement order of each information, thereby reducing the indication overhead to a certain extent.

[0119] The to-be-indicated information can be sent as a whole, or can be sent separately into multiple sub-information, and the sending period and / or sending occasion of the sub-information can be the same or different. The specific sending method is not limited by the present application. Wherein the sending period and / or sending occasion of the sub-information can be pre-defined, for example, pre-defined according to the protocol, or can be configured by the transmitting end device by sending configuration information to the receiving end device.

[0120] It can be understood that "sending" and "receiving" in the present application represent the direction of signal transmission. For example, "sending information to XX" can be understood as the destination of the information being XX, which can include direct transmission through the air interface, or indirect transmission through the air interface by other units or modules. "Receiving information from YY" can be understood as the source of the information being YY, which can include direct reception from YY through the air interface, or indirect reception from YY through the air interface by other units or modules. "Sending" can also be understood as the "output" of the chip interface, and "receiving" can also be understood as the "input" of the chip interface.

[0121] In other words, sending and receiving can be between devices, such as between network devices and terminal devices, or within devices, such as between components, modules, chips, software modules or hardware modules within a device through a bus, wire or interface.

[0122] It can be understood that the information may be processed as necessary between the source and destination of the information transmission, such as encoding, modulation, etc., but the destination can understand the valid information from the source. Similar expressions in the present application can be similarly understood and will not be repeated.

[0123] The technical solutions of the present application can be applied to a third generation partnership project (3rd generation partnership project, 3GPP) related cellular communication system. For example, a fourth generation (4th generation, 4G) communication system, a fifth generation (5th generation, 5G) communication system, a future communication system after the fifth generation communication system. For example, the fourth generation communication system can include a long term evolution (long term evolution, LTE) communication system. The fifth generation communication system can include a new radio (new radio, NR) communication system. The technical solutions of the present application can also be applied to a wireless fidelity (wireless fidelity, WiFi) system, a communication system supporting multiple wireless technology integration, a device-to-device (device-to-device, D2D) system, or a vehicle-to-everything (vehicle to everything, V2X) communication system, etc.

[0124] The communication system to which the technical solutions of the present application are applicable includes a first device and a second device. The first device and the second device can perform the technical solutions of the present application.

[0125] In a possible implementation, the first apparatus is a terminal device, or a chip, a chip system, or a processor in the terminal device, or a logic module or software for implementing part or all of the functions of the terminal device. The second apparatus is an access network device, or a chip, a chip system, or a processor in the access network device, or a logic module or software for implementing part or all of the functions of the access network device. For example, as shown in FIG. 2A, the first apparatus is a terminal device 201, and the second apparatus is an access network device 202.

[0126] In another possible implementation, the first apparatus is an access network device, or a chip, a chip system, or a processor in the access network device, or a logic module or software for implementing part or all of the functions of the access network device. The second apparatus is a terminal device, or a chip, a chip system, or a processor in the terminal device, or a logic module or software for implementing part or all of the functions of the terminal device. For example, as shown in FIG. 2A, the first apparatus is an access network device 202, and the second apparatus is a terminal device 201.

[0127] In yet another possible implementation, the first apparatus is an access network device, or a chip, a chip system, or a processor in the access network device, or a logic module or software for implementing part or all of the functions of the access network device. The second apparatus is a core network device, or a chip, a chip system, or a processor in the core network device, or a logic module or software for implementing part or all of the functions of the core network device. For example, as shown in FIG. 2A, the first apparatus is an access network device 202, and the second apparatus is a core network device 203.

[0128] In yet another possible implementation, the first apparatus is a core network device, or a chip, a chip system, or a processor in the core network device, or a logic module or software for implementing part or all of the functions of the core network device. The second apparatus is an access network device, or a chip, a chip system, or a processor in the access network device, or a logic module or software for implementing part or all of the functions of the access network device. For example, as shown in FIG. 2A, the first apparatus is a core network device 203, and the second apparatus is an access network device 202.

[0129] In yet another possible implementation, the first apparatus is a first terminal device, or a chip, a chip system, or a processor in the first terminal device, or a logic module or software for implementing part or all of the functions of the first terminal device. The second apparatus is a second terminal device, or a chip, a chip system, or a processor in the second terminal device, or a logic module or software for implementing part or all of the functions of the second terminal device. For example, as shown in FIG. 2B, the first apparatus is a terminal device 1, and the second apparatus is a terminal device 2.

[0130] The first device and the second device can also be other forms of devices, and the specific application does not make any limitation.

[0131] The terminal device, the access network device and the core network device involved in the present application are introduced as follows.

[0132] The terminal device can be a wireless terminal device capable of receiving scheduling information and indication information of the access network device. The wireless terminal device can be a device providing voice and / or data connectivity to a user, or a handheld device having a wireless connection function, or other processing devices connected to a wireless modem.

[0133] The terminal device can communicate with one or more core networks or the Internet via the access network. The terminal device can be a mobile terminal device, such as a mobile phone (or called "cellular" phone, mobile phone), a computer and a data card, for example, a portable, pocket-sized, handheld, computer-built-in or vehicle-mounted mobile device, which exchanges voice and / or data with a wireless access network. For example, personal communication service (PCS) phone, cordless phone, session initiation protocol phone, wireless local loop (WLL) station, personal digital assistant (PDA), tablet computer (Pad), computer with wireless transceiver function and the like. The wireless 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), customer premises equipment (CPE), a terminal, user equipment (UE), a mobile terminal (MT) and the like.

[0134] By way of example and not limitation, in embodiments of the present application, the terminal device can also be a wearable device. The wearable device can also be referred to as a smart wearable device or a smart wearable device, etc. It is a general term for devices that apply wearable technology to the intelligent design and development of daily wear. For example, glasses, gloves, watches, clothing, and shoes, etc. The wearable device is a portable device that can be directly worn on the body or integrated into the user's clothes or accessories. The wearable device is not just a hardware device, but also has powerful functions through software support and data interaction, cloud interaction. Broadly speaking, smart wearable devices include full-featured, large-sized devices that can achieve complete or partial functions without relying on smartphones, such as smartwatches or smartglasses, etc., and devices that focus on a specific application function and need to be used in conjunction with other devices such as smartphones, such as various smart wristbands, smart helmets, smart jewelry, etc.

[0135] The terminal device can also be a drone, a robot, a terminal device in device-to-device (D2D) communication, a terminal device in vehicle to everything (V2X), a virtual reality (VR) terminal device, an augmented reality (AR) terminal device, a wireless terminal device in industrial control, a wireless terminal device in self driving, a wireless terminal device in remote medical, a wireless terminal in smart grid, a wireless terminal device in transportation safety, a wireless terminal device in smart city, or a wireless terminal device in smart home, etc.

[0136] In addition, the terminal device can also be a terminal device in a future communication system evolved from the 5th generation (5G) communication system or a terminal device in a future evolved public land mobile network (PLMN), etc. For example, the future communication system can further expand the form and function of the 5G communication terminal. The terminal in the future communication system includes but is not limited to vehicles, cellular network terminals (with satellite terminal functions), drones, or internet of things (IoT) devices.

[0137] In the embodiments of the present application, the terminal device has an artificial intelligence (AI) capability. For example, the terminal device can obtain an AI service provided by a network device or a server. The terminal device also has an AI processing capability.

[0138] It should be noted that the terminal device can be a device or apparatus with a chip, or a device or apparatus integrated with a circuit, or a chip, a module or a control unit in the above-mentioned device or apparatus, and the specific embodiments of the present application are not limited.

[0139] The access network device can be a device in a wireless network. For example, the access network device can be an access network node (also referred to as a base station) that accesses a terminal device to a wireless network. Currently, some examples of the access network device are: a base station (gNodeB, 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 (for example, a home evolved Node B, or a home Node B, HNB), a baseband unit (BBU), or a wireless fidelity (Wi-Fi) access point (AP), and the like. In addition, in a network structure, the access network device can include a centralized unit (CU) node, a distributed unit (DU) node, a CU-control plane (CP), a CU-user plane (UP), or a radio unit (RU), or a RAN device including the CU node and the DU node. The CU and the DU can be separately arranged, or can be included in the same network element, such as a baseband unit (BBU). The RU can be included in a radio frequency device or a radio frequency unit, such as a remote radio unit (RRU), an active antenna unit (AAU), or a remote radio head (RRH). In different systems, the CU (or CU-CP and CU-UP), DU, or RU can also have different names, but those skilled in the art can understand their meanings. For example, in an open RAN (ORAN) system, the CU can also be referred to as an open CU (O-CU), the DU can also be referred to as an open DU (O-DU), the CU-CP can also be referred to as an open CU-CP (O-CU-CP), the CU-UP can also be referred to as an open CU-UP (O-CU-UP), and the RU can also be referred to as an open RU (O-RU). Any one of the CU (or CU-CP, CU-UP), DU, and RU can be implemented by a software module, a hardware module, or a combination of a software module and a hardware module.

[0140] The access network device can be other apparatuses providing wireless communication functions for the terminal device. Embodiments of the present application do not limit the specific technology and specific device form of the access network device. For the convenience of description, embodiments of the present application do not limit.

[0141] The core network device includes, for example, a mobility management entity (MME) in a fourth generation (4G) network, a home subscriber server (HSS), a serving gateway (S-GW), a policy and charging rules function (PCRF), a public data network gateway (P-GW), a network element 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 device can also include other core network devices in a 5G network and future networks of the 5G network. Optionally, the core network device is configured to be responsible for the mobile management of the terminal device and the like.

[0142] In embodiments of the present application, the access network device and the core network device described above can be network nodes with AI capabilities, and can provide AI services for terminal devices or other network devices. For example, the access network device and the core network device can be AI nodes, computing nodes, access network nodes with AI capabilities, or core network network elements with AI capabilities on the network side (access network or core network).

[0143] It should be noted that the access network device and the core network device can be the devices or apparatuses described above, or can be components (such as chips), modules, or units in the devices or apparatuses described above. The specific embodiments of the present application are not limited.

[0144] In a wireless communication system, AI technology can be used for compression and reconstruction of wireless channel information, beam management, and positioning enhancement, etc. An AI model is trained based on data to improve the performance of completing wireless tasks through the AI model. As shown in FIG. 1, a wireless AI framework includes data collection, model training, model management, model inference, and model storage. A device can measure an index of the AI model based on data to obtain the performance of the AI model. Optionally, the device performs management of the model according to the measured result. For example, switching of the AI model, or selection of the AI model, etc. As can be seen, the index measurement of the AI model is related to data, and when the data is incorrect or is abnormal data, the index of the AI model measured by the device is inaccurate. This further leads to unnecessary management of the AI model. For example, unnecessary switching, retraining, or adjustment of model parameters of the AI model, etc.

[0145] The present application provides a communication method, for a first device to measure an index of a first model according to first data, to determine a first index measurement result; and the first device to filter the first index measurement result to obtain a filtered index measurement result. This avoids inaccurate index measurement results, or in other words, improves the accuracy of the measured index measurement result, that is, more accurately reflects the index of the first model. This avoids unnecessary model management, and improves the efficiency of model management.

[0146] The technical solutions of the present application will be described below in conjunction with specific embodiments.

[0147] FIG. 3 is a schematic diagram of one embodiment of the communication method of the present application. Please refer to FIG. 3, the method includes:

[0148] 301, the first device measures an index of a first model according to first data, to determine a first index measurement result.

[0149] The first data is measurement data or monitoring data, and the first data is used to measure the inference performance or training performance of the first model. Optionally, the first index measurement result includes at least one of the following: a model index, or a system performance. The system performance refers to the system performance of a communication system in a case where the first device and a second device perform wireless related transmission using the first model. The model index includes at least one of the following: accuracy of a model inference result of the first model, loss function value of model training, change amount of model parameters of model training, performance of a validation data set, or input / output distribution drift.

[0150] For example, the first data comprises t beam IDs of the first device at t time instants, where t is an integer greater than or equal to 2, and the t beam IDs are obtained by the conventional beam sweeping. The first device uses the beam IDs of the first device at m time instants as input data of the first model, and the first model outputs a predicted beam ID of the first device at m+1 time instant. m is an integer greater than or equal to 1. The first device compares the predicted beam ID of the first device at m+1 time instant by the first model with the beam ID of the first device at m+1 time instant included in the first data, and obtains a comparison result. The comparison process for the beam IDs of the first device at other time instants is similar. Therefore, the first device can obtain t-1 comparison results. The first device determines a model indicator of the first model according to the t-1 comparison results. When the t-1 comparison results comprise a comparison result of a pair of beam IDs, the first device can determine the reliability of the predicted beam ID by the first model. For example, the first device calculates a cross-entropy, which is used to indicate the difference between the predicted beam by the first model and the beam determined by the conventional beam sweeping. When the t-1 comparison results comprise comparison results of multiple pairs of beam IDs, the first device can determine the beam prediction accuracy of the first model. The beam prediction accuracy comprises the number of predicted accurate beams over the total number of beams predicted by the first model.

[0151] For another example, the first device performs data transmission by the predicted beam by the first model. In the case that the first device performs data transmission by the predicted beam by the first model, the first device determines a system performance of the communication system. For example, the success decoding probability (e.g., block error rate (BLER)) or the transmission rate of the data transmission, etc.

[0152] Optionally, the embodiment shown in FIG. 3 further comprises step 301a. Step 301a can be performed before step 301.

[0153] 301a. The second device sends second configuration information to the first device. Correspondingly, the first device receives the second configuration information from the second device.

[0154] The second configuration information is used to indicate a measurement configuration. Optionally, the measurement configuration comprises at least one of the following: an indicator to be measured, or a number of samples used to determine the indicator to be measured. Optionally, the indicator to be measured comprises at least one of the following: a model indicator, or a system performance. The model indicator comprises at least one of the following: the accuracy of a model inference result, a loss function value of model training, a variation of a model parameter of model training, a validation dataset performance, or an input-output distribution drift.

[0155] Optionally, the number of sample data used to determine the indicator to be measured is one or more.

[0156] For example, for the CSI compression scenario, the sample data can be predicted channel information. For example, the predicted channel information includes frequency domain channels or precoding matrices of F subbands on K antennas. For example, the frequency domain channels of F subbands on K antennas are one sample data, and K and F are both integers greater than or equal to 1, or the precoding matrix is one sample data. For the case of one sample data, it is referred to as one piece of predicted channel information here. For the case of multiple sample data, it is referred to as multiple pieces of predicted channel information here. If the sample data includes one piece of predicted channel information, the first device can calculate the normalized mean square error (NMSE) or the square generalized cosine similarity (SGCS) of the predicted channel information and the channel information obtained through channel estimation. If the sample data includes multiple pieces of predicted channel information, the first device can calculate the average NMSE or the average SGCS according to the multiple pieces of predicted channel information and the multiple pieces of channel information obtained through channel estimation. For another example, for the beam prediction scenario, the sample data can be predicted beam IDs. If the sample data is one predicted beam ID, the first device can calculate the reliability of the predicted beam ID. If the sample data is multiple predicted beam IDs, that is, the sample data includes multiple beam IDs, the first device can calculate the accuracy or the correct probability of the multiple predicted beam IDs.

[0157] Optionally, when the number of sample data used to determine the to-be-measured index is multiple, the multiple sample data can be obtained at different times, at different positions, by different beams, different resource elements (REs), and / or different terminal devices.

[0158] Optionally, the step 301 specifically includes: determining, by the first device, the first index measurement result according to the measurement configuration.

[0159] 302. The first device filters the first index measurement result to obtain a filtered index measurement result.

[0160] Optionally, the filtering in the step 302 can also be described as processing, or filtering processing, etc., and the specific embodiments are not limited herein. Optionally, the filtering includes that the first device performs weighted processing on the first index measurement result and the historical index measurement result or the historical filtered index measurement result of the first model. For example, as shown in FIG. 4, the first device measures the index of the first model to obtain the first index measurement result. Then, the first device filters the first index measurement result to obtain the filtered index measurement result.

[0161] Optionally, the first device filters the measurement result of the first index according to the target filtering method to obtain the filtered index measurement result.

[0162] The target filtering method is either pre-configured, determined by the first device, or configured by the second device. Optionally, the embodiment shown in Figure 3 further includes step 301b. Step 301b can be performed before step 302.

[0163] 301b. The second device sends first configuration information to the first device. Correspondingly, the first device receives the first configuration information from the second device.

[0164] The first configuration information is used to indicate the target filtering method.

[0165] The following describes some possible implementations of the target filtering method. Other implementations are still applicable to this application, and this application does not limit them.

[0166] I. The target filtering method is the first filtering method. The first filtering method includes weighting the measurement results of the first index with the historical filtered index measurement results. The historical filtered index measurement results include the filtered index measurement results obtained by the first device measuring the index of the first model before the first index measurement results and then filtering it.

[0167] Optionally, the aforementioned first configuration information includes a first weighting factor. The first weighting factor is used to weight the first indicator measurement result with the historical filtered indicator measurement result. For example, as shown in Figure 5, the filtered indicator measurement result is represented as P. n Then P n =(1-a)P n-1 +a*M n Where a is the first weighting factor, P1 = M1, P2 = (1-a)P1 + aM2, ..., P n-1 =(1-a)P n-2 +aM n-1 M1 represents the index measurement result obtained by the first device at the first time or at the first moment when measuring the index of the first model. M2 represents the index measurement result obtained by the first device at the second time or at the second moment when measuring the index of the first model. P1 represents the filtered index measurement result obtained by the first device filtering M1. Since the first device only has an index measurement result at one time or at one moment, P1 = M1. P2 represents the filtered index measurement result obtained by the first device filtering M2. n-1 The index measurement result is obtained by measuring the index of the first model by the first device at the (n-1)th time or the (n-1)th time. n-1 For the first device to M n-1 The filtered index measurement result obtained after filtering. Mn is a first index measurement result, i.e., an index measurement result obtained by the first device measuring an index of the first model at the nth time or at the nth time point. n-1 is a historical filtered index measurement result, i.e., a filtered index measurement result obtained by the first device filtering M n-1 is an integer greater than or equal to 2. It should be noted that the first weighting factor can also be preconfigured or determined by the first device, which is not limited in the present application.

[0168] Optionally, the first weighting factor can be 1 / 4, 1 / 8, 1 / 16, 1 / 32, or 3 / 16.

[0169] Optionally, the first weighting factor can be determined according to a signal-to-noise ratio adopted by the first device when measuring the index of the first model. The greater the signal-to-noise ratio, the smaller the first weighting factor.

[0170] Optionally, the first filtering manner can also be referred to as continuous filtering.

[0171] II. The target filtering manner is a second filtering manner, and the second filtering manner comprises weighting the first index measurement result and the historical index measurement result.

[0172] The first index measurement result is obtained by the first device measuring the index of the first model at the Nth time or at the Nth time point. The historical index measurement result comprises index measurement results obtained by the first device measuring the index of the first model at the K previous time points of the Nth time, or measuring the index of the first model at the K previous times of the Nth time. K is an integer greater than or equal to 1, and N is an integer greater than or equal to 2. Optionally, the second filtering manner comprises averaging the first index measurement result and the historical index measurement result. For example, as shown in FIG. 6, the filtered index measurement result is represented as P n , then P n = 1 / K (M n-K+1 +…+M n ), wherein M n-K+1 is an index measurement result obtained by the first device measuring the index of the first model at the N-K+1th time or at the N-K+1th time point. M n is the first index measurement result.

[0173] Optionally, the second filtering manner can also be referred to as sliding window filtering. The above K can be understood as a window length of the sliding window filtering.

[0174] Optionally, the above first configuration information comprises at least one second weighting factor and / or K. The at least one second weighting factor is used for weighting the first index measurement result and the historical index measurement result.

[0175] The at least one second weighting factor comprises a second weighting factor corresponding to the first index measurement result and a second weighting factor corresponding to an index measurement result obtained by the first device measuring the index of the first model at each time in the K previous times of the Nth time, or a second weighting factor corresponding to the first index measurement result and a second weighting factor corresponding to an index measurement result obtained by the first device measuring the index of the first model at each time in the K previous times of the Nth time. For example, the at least one second weighting factor comprises a second weighting factor a1, a second weighting factor a2, and a second weighting factor a3. The second weighting factor a1 is a second weighting factor corresponding to the first index measurement result. The second weighting factor a2 is a second weighting factor corresponding to an index measurement result obtained by the first device measuring the index of the first model at the N-1th time. The second weighting factor a3 is a second weighting factor corresponding to an index measurement result obtained by the first device measuring the index of the first model at the N-2th time. Therefore, the filtered index measurement result P n = a1M n + a2M n-1 + a3M n-2 , where M n is the first index measurement result, M n-1 is an index measurement result obtained by the first device measuring the index of the first model at the N-1th time, and M n-2 is an index measurement result obtained by the first device measuring the index of the first model at the N-2th time.

[0176] It should be noted that the at least one second weighting factor and / or K can also be preconfigured or determined by the first device, which is not limited in the present application.

[0177] Optionally, each second weighting factor can be determined according to a time interval between the time corresponding to the index measurement result corresponding to the second weighting factor and the Nth time. For example, the larger the time interval, the smaller the second weighting factor.

[0178] Optionally, each second weighting factor can be determined according to an interval number between the number corresponding to the index measurement result corresponding to the second weighting factor and the Nth time. For example, the larger the interval number, the smaller the second weighting factor.

[0179] It should be noted that there is no fixed execution order between step 301b and step 301. For example, step 301b is executed first, and then step 301 is executed; or step 301 is executed first, and then step 301b is executed; or step 301b and step 301 are executed simultaneously according to the situation, which is not limited in the present application.

[0180] Optionally, if the embodiment shown in FIG. 3 further includes step 301a, there is no fixed execution order between step 301a and step 301b. Step 301a can be executed first, and then step 301b can be executed; or step 301b can be executed first, and then step 301a can be executed; or step 301a and step 301b can be executed at the same time according to the situation, which is not limited in the present application.

[0181] Optionally, the first configuration information and the second configuration information are the same configuration information.

[0182] Optionally, the first device can also perform anomaly detection on the first data. The following introduces a possible implementation manner of step 301. Optionally, step 301 specifically includes steps 1 to 3.

[0183] Step 1: The first device measures an index of the first model according to the first data, and obtains a second index measurement result.

[0184] Step 2: The first device performs anomaly detection on the first data, and obtains abnormal data.

[0185] It should be noted that there is no fixed execution order between step 1 and step 2. Step 1 can be executed first, and then step 2 can be executed; or step 2 can be executed first, and then step 1 can be executed; or step 1 and step 2 can be executed at the same time according to the situation, which is not limited in the present application.

[0186] The following introduces some possible implementation manners of the abnormal data. The present application is still applicable to other implementation manners, which is not limited in the present application.

[0187] Implementation manner one: The first data includes one or more data, and the abnormal data includes data whose value difference with a reference data value is greater than or equal to a first threshold value. Or, the abnormal data includes data whose absolute value of the value difference with the reference data value is greater than or equal to the first threshold value. Or, the abnormal data includes data whose value difference with the reference data value is greater than the first threshold value. Or, the abnormal data includes data whose absolute value of the value difference with the reference data value is greater than the first threshold value.

[0188] In a possible implementation manner, the first data includes multiple data, and the reference data value is a weighted value or a median value of the multiple data. The weighted value of the multiple data is a weighted value obtained by weighting the values of the multiple data. The median value of the multiple data is a value in the middle of the values of the multiple data.

[0189] In another possible implementation, the reference data value is a weighted value or a median value of historical data. The historical data is historical data of the index used to measure the first module. The weighted value of the historical data is a weighted value obtained by weighting the historical data. The median value of the historical data is a value that is in the middle of values of the historical data.

[0190] Optionally, the first threshold value is equal to X times a first standard deviation, the first standard deviation being a standard deviation of the plurality of data included in the first data. X is greater than 0. For example, X = 3. Of course, the first threshold value can also be other values, which are not limited in the present application.

[0191] It should be noted that the first threshold value can be preconfigured, or indicated by the second device to the first device, or determined by the first device, which are not limited in the present application.

[0192] Implementation two: the second index measurement result includes one or more model indexes. The abnormal data is data corresponding to the index measurement result, a difference between the one or more model indexes and a reference model index being greater than or equal to a second threshold value. Alternatively, the abnormal data is data corresponding to the index measurement result, an absolute value of the difference between the one or more model indexes and the reference model index being greater than or equal to the second threshold value. Alternatively, the abnormal data is data corresponding to the index measurement result, the difference between the one or more model indexes and the reference model index being greater than the second threshold value. Alternatively, the abnormal data is data corresponding to the index measurement result, the absolute value of the difference between the one or more model indexes and the reference model index being greater than the second threshold value.

[0193] Specifically, the one or more model indexes are indexes of the first model measured by the first device according to the first data. Therefore, each model index corresponds to part or all of the data in the first data.

[0194] In a possible implementation, the second index measurement result includes a plurality of model indexes, and the reference model index is obtained by weighting the plurality of model indexes, or is a model index in the middle of the plurality of model indexes.

[0195] In another possible implementation, the reference model index is obtained by weighting historical model indexes, or is a model index in the middle of the historical model indexes. The historical model indexes are model indexes measured by measuring indexes of the first model before the first index measurement result.

[0196] Optionally, the second threshold value is equal to Y times a second standard deviation, the second standard deviation being a standard deviation of the plurality of model indexes. Y is greater than 0. For example, Y = 3. Of course, the second threshold value can also be other values, which are not limited in the present application.

[0197] It should be noted that the second threshold value may be pre-configured, indicated by the second device to the first device, or determined by the first device; this application does not specify the specific value.

[0198] Implementation Method 3: The second indicator measurement result includes the performance of one or more systems. Abnormal data is data corresponding to indicator measurement results where the difference between the performance of the one or more systems and the performance of the reference system is greater than or equal to a third threshold. Alternatively, abnormal data is data corresponding to indicator measurement results where the absolute value of the difference between the performance of the one or more systems and the performance of the reference system is greater than or equal to a third threshold. Alternatively, abnormal data is data corresponding to indicator measurement results where the difference between the performance of the one or more systems and the performance of the reference system is greater than a third threshold. Alternatively, abnormal data is data corresponding to indicator measurement results where the absolute value of the difference between the performance of the one or more systems and the performance of the reference system is greater than a third threshold.

[0199] Specifically, the performance of one or more systems is obtained by the first device measuring the indicators of the first model based on the first data. Therefore, each system performance corresponds to some or all of the data in the first data.

[0200] In one possible implementation, the second metric measurement result includes multiple system performance metrics, and the reference system performance is obtained by weighting these multiple system performance metrics, or it is the system performance metric that is in the middle among the multiple system performance metrics.

[0201] In another possible implementation, the reference system performance is obtained by weighting historical system performance, or by using a system performance that is intermediate in magnitude from the historical system performance. The historical system performance is the system performance obtained by measuring the first model's metrics before the first metric measurement results.

[0202] Step 3: The first device determines the measurement result of the first index.

[0203] The first indicator measurement result is the indicator measurement result in the second indicator measurement result excluding the indicator measurement results corresponding to abnormal data. Specifically, the second indicator measurement result is obtained by the first device measuring the indicators of the first model based on the first data. Therefore, each indicator measurement result corresponds to part or all of the data in the first data. The first indicator measurement result is the remaining indicator measurement result in the second indicator measurement result after excluding the indicator measurement results corresponding to abnormal data. For example, as shown in Figure 7, the first device measures the indicators of the first model to obtain the second indicator measurement result. Then, the first device performs anomaly filtering or anomaly detection on the first data based on the second indicator measurement result to obtain abnormal data. Then, the first device determines the first indicator measurement result in the second indicator measurement result excluding the indicator measurement results corresponding to abnormal data. The first device then filters the first indicator measurement result.

[0204] In a possible implementation, the second index measurement result includes one or more model indexes. The index measurement result corresponding to the abnormal data is an index measurement result whose difference with the reference model index is greater than or equal to the second threshold value, among the one or more model indexes. Alternatively, the index measurement result corresponding to the abnormal data is an index measurement result whose absolute value of the difference with the reference model index is greater than or equal to the second threshold value, among the one or more model indexes. Alternatively, the index measurement result corresponding to the abnormal data is an index measurement result whose difference with the reference model index is greater than the second threshold value, among the one or more model indexes. Alternatively, the index measurement result corresponding to the abnormal data is an index measurement result whose absolute value of the difference with the reference model index is greater than the second threshold value, among the one or more model indexes. For the reference model index and the second threshold value, refer to the foregoing relevant description, which is not repeated here.

[0205] In another possible implementation, the second index measurement result includes one or more system performances. The index measurement result corresponding to the abnormal data is an index measurement result whose difference with the reference system performance is greater than or equal to the third threshold value, among the one or more system performances. Alternatively, the index measurement result corresponding to the abnormal data is an index measurement result whose absolute value of the difference with the reference system performance is greater than or equal to the third threshold value, among the one or more system performances. Alternatively, the index measurement result corresponding to the abnormal data is an index measurement result whose difference with the reference system performance is greater than the third threshold value, among the one or more system performances. Alternatively, the index measurement result corresponding to the abnormal data is an index measurement result whose absolute value of the difference with the reference system performance is greater than the third threshold value, among the one or more system performances. For the reference system performance and the third threshold value, refer to the foregoing relevant description, which is not repeated here.

[0206] It should be noted that the foregoing is an example in which the first device performs abnormality detection on the first data, and measures indexes of the first model according to the first data. In actual application, if the first device has first data abnormality, the first device can not measure indexes of the first model, that is, the first device can ignore the first data.

[0207] Optionally, the first device can send the filtered index measurement result to the second device, to facilitate the second device in managing the first model. In this implementation, the embodiment shown in FIG. 3 further includes steps 303 to 304. Steps 303 to 304 can be performed after step 302.

[0208] Optionally, in step 302, the first device performs the anomaly detection on the abnormal data to obtain the abnormal data. Optionally, the embodiment shown in FIG. 3 further includes step 303a. Step 303a can be performed after step 302.

[0209] 303a, the first device sends the anomaly result to the second device. Correspondingly, the second device receives the anomaly result from the first device.

[0210] The anomaly result includes the abnormal data, the number of anomalies, and / or the proportion of the abnormal data.

[0211] Optionally, the second index measurement result is obtained by the first device according to the first data. The number of anomalies is the number of times that the index of the first model obtained by the first device through multiple measurements is abnormal. The proportion of the abnormal data is the ratio of the abnormal data to the first data.

[0212] 303, the first device sends the filtered index measurement result to the second device. Correspondingly, the second device receives the filtered index measurement result from the first device.

[0213] It should be noted that if the embodiment shown in FIG. 3 further includes step 303a, there is no fixed execution order between step 303 and step 303a. Step 303 can be executed first, and then step 303a can be executed. Alternatively, step 303a can be executed first, and then step 303 can be executed. Alternatively, step 303a and step 303 can be executed simultaneously according to the situation, and the specific application is not limited.

[0214] Optionally, the embodiment shown in FIG. 3 further includes step 301c. Step 301c can be performed before step 303.

[0215] 301c, the second device sends the third configuration information to the first device. Correspondingly, the first device receives the third configuration information from the second device.

[0216] The third configuration information is used to indicate the reporting configuration. Optionally, the reporting configuration includes at least one of the following: a reporting condition or a reporting period. For example, the reporting condition includes: reporting the filtered index measurement result when the index of the first model in the filtered index measurement result is greater than or equal to a corresponding threshold value; or reporting the filtered index measurement result when the index of the first model in the filtered index measurement result is greater than a corresponding threshold value; or reporting the filtered index measurement result when the index of the first model in the filtered index measurement result is less than or equal to a corresponding threshold value; or reporting the filtered index measurement result when the index of the first model in the filtered index measurement result is less than a corresponding threshold value.

[0217] Optionally, the reporting period is the reporting period of the filtered index measurement result when the reporting condition is met.

[0218] Optionally, the step 303 specifically comprises: the first device sending the filtered index measurement result to the second device according to the reporting configuration.

[0219] It should be noted that there is no fixed execution order between the step 301c and the steps 301-302. The step 301c can be executed first, and then the steps 301-302 can be executed; or the steps 301-302 can be executed first, and then the step 301c can be executed; or the step 301c and the steps 301-302 can be executed simultaneously according to the situation, and the specific application is not limited.

[0220] Optionally, if the embodiment shown in FIG. 3 further comprises the step 303a, there is no fixed execution order between the step 303a and the step 301c. The step 303a can be executed first, and then the step 301c can be executed; or the step 301c can be executed first, and then the step 303a can be executed; or the step 303a and the step 301c can be executed simultaneously according to the situation, and the specific application is not limited.

[0221] Optionally, if the embodiment shown in FIG. 3 further comprises the step 301a, there is no fixed execution order between the step 301a and the step 301c. The step 301a can be executed first, and then the step 301c can be executed; or the step 301c can be executed first, and then the step 301a can be executed; or the step 301a and the step 301c can be executed simultaneously according to the situation, and the specific application is not limited. Optionally, the second configuration information and the third configuration information are the same configuration information.

[0222] Optionally, if the embodiment shown in FIG. 3 further comprises the step 301b, there is no fixed execution order between the step 301b and the step 301c. The step 301b can be executed first, and then the step 301c can be executed; or the step 301c can be executed first, and then the step 301b can be executed; or the step 301b and the step 301c can be executed simultaneously according to the situation, and the specific application is not limited. Optionally, the first configuration information and the third configuration information are the same configuration information.

[0223] Optionally, if the embodiment shown in FIG. 3 further comprises the step 301a and the step 301b, there is no fixed execution order among the step 301a, the step 301b and the step 301c. For example, the step 301a can be executed first, then the step 301b can be executed, and finally the step 301c can be executed, and the specific application is not limited. Optionally, the first configuration information, the second configuration information and the third configuration information are the same configuration information.

[0224] 304、The second device manages the first model according to the filtered index measurement result.

[0225] Optionally, the second device performs inference management or training management on the first model according to the filtered index measurement result.

[0226] Optionally, the filtered index measurement result comprises a model index and / or system performance of the first model, and the inference management comprises any one of the following:

[0227] I. If the model index and / or system performance is less than or equal to a fourth threshold value, the second device falls back to a default mode, wherein the default mode is to perform transmission in a non-AI manner or by using a default AI model, or to measure an index of a third model, or to switch to a fourth model. Alternatively, if the model index and / or system performance is less than the fourth threshold value, the second device falls back to a default mode, wherein the default mode is to perform transmission in a non-AI manner or by using a default AI model, or to measure an index of a third model, or to switch to a fourth model.

[0228] For example, if the first model is an activated model and the model index and / or system performance is less than or equal to the fourth threshold value, the second device can fall back to a default mode, or measure an index of a third model, or switch to a fourth model. For example, the second device performs transmission by using a predicted beam output by the first model, and the default mode can be that the second device determines a beam by using a conventional beam sweeping manner and performs transmission through the beam.

[0229] Optionally, the fourth threshold value can be determined according to system performance when transmission is performed in a non-AI manner; or the fourth threshold value can be determined according to a model index or system performance when transmission is performed by using a default AI model; or the fourth threshold value can be determined according to a model index or system performance when transmission is performed by using a fourth model.

[0230] II. If the model index and / or system performance is greater than or equal to a fifth threshold value, the first model is activated, or the second device switches to the first model. Alternatively, if the model index and / or system performance is greater than the fifth threshold value, the first model is activated, or the second device switches to the first model.

[0231] For example, if the first model is a non-activated model and the model index and / or system performance is greater than or equal to the fifth threshold value, the first model is activated, or the second device switches to the first model.

[0232] Optionally, the fifth threshold value can be determined according to system performance when transmission is performed in a non-AI manner; or the fifth threshold value can be determined according to a model index or system performance when transmission is performed by using a default AI model; or the fifth threshold value can be determined according to a model index or system performance when transmission is performed by using a currently activated model.

[0233] III. If the model index and / or system performance of the first model measured by the first device for Q consecutive times is less than or equal to the sixth threshold value, the first device requests to train the first model, or requests to collect a data set for training the first model, where Q is an integer greater than or equal to 1. Alternatively, if the model index and / or system performance of the first model measured by the first device for Q consecutive times is less than the sixth threshold value, the first device requests to train the first model, or requests to collect a data set for training the first model, where Q is an integer greater than or equal to 1.

[0234] Optionally, the value of Q can be determined according to the use scenario or use range of the first model. For example, if the first model supports fast adjustment, the value of Q can be small, for example, Q = 10 or 100. If the first model does not support fast adjustment, the value of Q can be large. For example, Q = 1000.

[0235] Optionally, the sixth threshold value can be determined according to the system performance when the transmission is performed in a non-AI manner, or the sixth threshold value can be determined according to the model index or system performance when the transmission is performed by using a default AI model.

[0236] Optionally, the filtered index measurement result includes the model index and / or system performance of the first model. The following describes some possible implementation manners of the training management. The present application is still applicable to other implementation manners.

[0237] I. If the model index and / or system performance is less than or equal to the seventh threshold value, the training of the first model is terminated, or the learning rate of the first model is adjusted. Alternatively, if the model index and / or system performance is less than the seventh threshold value, the training of the first model is terminated, or the learning rate of the first model is adjusted.

[0238] For example, if the filtered index measurement result includes the loss function value of the model training or the change amount of the model parameter of the model training, and the loss function value of the model training or the change amount of the model parameter of the model training is less than or equal to the seventh threshold value, the training of the first model is terminated, or the learning rate of the first model is adjusted. For example, the learning rate of the first model is reduced.

[0239] Optionally, the seventh threshold value can be determined according to the system performance when the transmission is performed in a non-AI manner, or the seventh threshold value can be determined according to the model index or system performance when the transmission is performed by using a default AI model, or the seventh threshold value can be determined according to the historical model index or historical system performance obtained from the historical training process of the first model. For example, the average value of the loss function value or the average value of the change amount of the model parameter in the historical training process of the first model.

[0240] If the model index and / or the system performance is greater than or equal to the eighth threshold value, the first model is initialized or the second model is loaded, the first model is obtained by training the AI service based on the first data for M rounds, and the second model is a model obtained by training the AI service for M-1 rounds, M being an integer greater than or equal to 2. Alternatively, if the model index and / or the system performance is greater than the eighth threshold value, the first model is initialized or the second model is loaded, the first model is obtained by training the AI service based on the first data for M rounds, and the second model is a model obtained by training the AI service for M-1 rounds, M being an integer greater than or equal to 2.

[0241] For example, the filtered index measurement result includes a loss function value of model training or a change amount of a model parameter of model training, and the loss function value of model training or the change amount of the model parameter of model training is greater than or equal to the eighth threshold value, the training of the first model is terminated, or the first model is initialized, or the second model is loaded.

[0242] Optionally, the eighth threshold value can be determined according to a historical model index or a historical system performance obtained in a historical training process of the first model. For example, an average value of a loss function value or an average value of a change amount of a model parameter in the historical training process of the first model.

[0243] It should be noted that the inference management or the training management can also be performed by the first device, and the specific application is not limited.

[0244] Optionally, after the second device receives the abnormal result and the filtered index measurement result, the second device can load the second model, or the second device updates the model parameter of the first model. For the second model, please refer to the related description in the foregoing. It should be noted that the implementation mode of loading the second model or updating the model parameter of the first model by the second device is taken as an example, and in actual application, the first device can also perform the implementation mode of loading the second model or updating the model parameter of the first model, and the specific application is not limited. That is, the second device determines the management operation of the model and instructs the first device to perform the management operation.

[0245] Optionally, the embodiment shown in FIG. 3 further includes step 305. Step 305 can be performed after step 304.

[0246] 305. The second device sends a management instruction to the first device. Correspondingly, the first device receives the management instruction from the second device.

[0247] The management instruction is used to instruct the first device to perform a corresponding management operation. For example, the management operation includes inference management or training management. Correspondingly, the first device performs the corresponding management operation according to the management instruction. For the inference management or the training management, please refer to the related description in the foregoing, which will not be repeated here.

[0248] Optionally, the embodiment shown in FIG. 3 further comprises steps 306 to 307. Steps 306 to 307 can be executed after step 302.

[0249] 306. The first device manages the first model according to the filtered index measurement result.

[0250] Optionally, the first device inferences manages or trains manages the first model according to the filtered index measurement result. For inference management or training management, please refer to the foregoing relevant introduction, which will not be repeated here.

[0251] Optionally, after the first device receives the abnormal result and the filtered index measurement result, the first device can load the second model, or the first device updates the model parameters of the first model. For the second model, please refer to the foregoing relevant introduction. It should be noted that here is an example of the first device performing loading the second model or updating the model parameters of the first model. In actual application, the second device can also perform loading the second model or updating the model parameters of the first model, and the specific application of the present application is not limited. That is, the first device determines the management operation of the model and instructs the second device to perform the management operation.

[0252] 307. The first device sends the management result of the first model to the second device. Correspondingly, the second device receives the management result of the first model from the first device.

[0253] The management result of the first model is used to indicate the inference management or training management of the first model by the first device.

[0254] Optionally, the embodiment shown in FIG. 3 further comprises step 308.

[0255] 308. The second device sends the management confirmation to the first device. Correspondingly, the first device receives the management confirmation from the second device.

[0256] The management confirmation is used to determine the management result of the first model. For example, the first device is a terminal device, and the second device is an access network device. The terminal device decides to switch from the first model to another model. Then, the terminal device sends the management result of the first model to the access network device, and the management result of the first model is used to indicate switching from the first model to another model. The access network device sends the management confirmation to the terminal device, and the management confirmation is used to determine switching from the first model to another model. Then, the terminal device receives the management confirmation, and the terminal device switches from the first model to another model.

[0257] In the embodiments of the present application, the first device determines the first index measurement result according to the first data measurement of the index of the first model. Then, the first device filters the first index measurement result to obtain a filtered index measurement result. The first device sends the filtered index measurement result to the second device, or the first device manages the first model according to the filtered index measurement result and sends the management result of the first model to the second device. As can be seen, the first device filters the first index measurement result to obtain the filtered index measurement result. Thus, it is avoided that the measured index measurement result is inaccurate, or in other words, the accuracy of the measured index measurement result is improved, that is, the index of the first model is more accurately reflected. Then, the first device manages the first model according to the filtered index measurement result, and efficient management of the first model is realized. Or, the first device sends the filtered index measurement result to the second device, which is convenient for the second device to manage the first model efficiently in combination with the filtered index measurement result. This is beneficial to avoid unnecessary model management and improve the efficiency of model management.

[0258] FIG. 8 is a structural schematic diagram of the first device according to an embodiment of the present application. Referring to FIG. 8, the first device can be used to execute the process executed by the first device in the embodiment shown in FIG. 3. For details, please refer to the related description in the above method embodiments.

[0259] The first device 800 includes a transceiver module 801 and a processing module 802.

[0260] The processing module 802 is configured to perform data processing. The transceiver module 801 can realize the corresponding communication function. The transceiver module 801 can also be referred to as a communication interface or a communication module.

[0261] Optionally, the first device 800 can further include a storage module, which can be used to store instructions and / or data. The processing module 802 can read the instructions and / or data in the storage module, so that the first device realizes the foregoing method embodiments.

[0262] The first device 800 can be used to execute the actions performed by the first device in the above method embodiments. The first device 800 can be the first device or a component configurable to the first device. The processing module 802 is configured to perform the processing-related operations of the first device side in the above method embodiments. The transceiver module 801 is configured to perform the receiving-related operations of the first device side in the above method embodiments.

[0263] Optionally, the transceiver module 801 can include a sending module and a receiving module. The sending module is configured to perform the sending operations in the above method embodiments. The receiving module is configured to perform the receiving operations in the above method embodiments.

[0264] It should be noted that the first device 800 can include a sending module but not a receiving module. Alternatively, the first device 800 can include a receiving module but not a sending module. Whether the first device 800 includes a sending module or a receiving module can depend on whether the first device 800 performs the sending action and the receiving action in the above-mentioned schemes.

[0265] Optionally, the first device 800 is configured to perform the actions performed by the first device in the above-mentioned embodiment shown in FIG. 3. For details, refer to the related description in the above-mentioned embodiment shown in FIG. 3, which will not be repeated here. For example, the first device 800 is configured to perform the following scheme.

[0266] The processing module 802 is configured to measure an index of the first model according to the first data, to obtain a first index measurement result; filter the first index measurement result to obtain a filtered index measurement result; and the transceiver module 801 is configured to send the filtered index measurement result to the second device. Alternatively, the processing module 802 is further configured to manage the first model according to the filtered index measurement result; and the transceiver module 801 is further configured to send a management result of the first model to the second device.

[0267] In a possible implementation, the transceiver module 801 is further configured to receive first configuration information from the second device, the first configuration information being used to indicate a target filtering manner; and the processing module 802 is specifically configured to filter the first index measurement result according to the target filtering manner to obtain the filtered index measurement result.

[0268] In another possible implementation, the target filtering manner is a first filtering manner, and the first filtering manner includes performing weighted processing on the first index measurement result and a historical filtered index measurement result, the historical filtered index measurement result including a filtered index measurement result obtained by filtering an index of the first model measured by the first device 800 before the first index measurement result. Alternatively, the target filtering manner is a second filtering manner, and the second filtering manner includes performing weighted processing on the first index measurement result and a historical index measurement result by the first device 800, the first index measurement result being an index measurement result obtained by measuring the index of the first model by the first device 800 at the Nth time or the Nth time, the historical index measurement result including an index measurement result obtained by measuring the index of the first model by the first device 800 at the Kth time before the Nth time, or an index measurement result obtained by measuring the index of the first model by the first device 800 at the Kth time before the Nth time, K being an integer greater than or equal to 1, and N being an integer greater than or equal to 2.

[0269] In a possible implementation, the first configuration information comprises a first weighting factor used for weighting the first indicator measurement result and the historical filtered indicator measurement result; or the first configuration information comprises at least one second weighting factor used for weighting the first indicator measurement result and the historical indicator measurement result, and / or K.

[0270] In a possible implementation, the transceiver 801 is further configured to receive second configuration information from the second device, the second configuration information being used for indicating a measurement configuration; and the processor 802 is specifically configured to measure the indicators of the first model according to the first data and the measurement configuration, and determine the first indicator measurement result.

[0271] In a possible implementation, the measurement configuration comprises at least one of the following: an indicator to be measured, or a sample number used for determining the indicator to be measured.

[0272] In a possible implementation, the indicator to be measured comprises at least one of the following: a model indicator, or a system performance, and the model indicator comprises at least one of the following: an accuracy of a model inference result, a loss function value of model training, a variation of a model parameter of model training, a validation data set performance, or an input / output distribution drift.

[0273] In a possible implementation, the processor 802 is specifically configured to: measure indicators of the first model according to the first data, to obtain a second indicator measurement result; perform anomaly detection on the first data according to the second indicator measurement result, to obtain anomaly data; and determine the first indicator measurement result, the first indicator measurement result being the indicator measurement result in the second indicator measurement result except for the indicator measurement result corresponding to the anomaly data.

[0274] In a possible implementation, the first data comprises one or more data, and the anomaly data comprises data whose value is greater than or equal to a first threshold value with reference to a reference data value.

[0275] In a possible implementation, the second indicator measurement result comprises one or more model indicators, and the indicator measurement result corresponding to the anomaly data is an indicator measurement result whose difference with a reference model indicator is greater than or equal to a second threshold value; or the second indicator measurement result comprises one or more system performances, and the indicator measurement result corresponding to the anomaly data is an indicator measurement result whose difference with a reference system performance is greater than or equal to a third threshold value.

[0276] In a possible implementation, the transceiver 801 is further configured to send, to the second device, an anomaly result, the anomaly result comprising the anomaly data, and / or a proportion of the anomaly data.

[0277] In another possible implementation, the processing module 802 is specifically configured to: load a second model, the first model being obtained by training the AI service based on the first data for M rounds, the second model being a model obtained by training the AI service for M-1 rounds, M being an integer greater than or equal to 2; or update the model parameters of the first model.

[0278] In another possible implementation, the transceiver module 801 is further configured to: receive third configuration information from the second device, the third configuration information being used to indicate a reporting configuration; and send the filtered indicator measurement result to the second device according to the reporting configuration.

[0279] In another possible implementation, the reporting configuration includes at least one of the following: a reporting condition or a reporting period.

[0280] In another possible implementation, the processing module 802 is specifically configured to: perform inference management or training management on the first model according to the filtered indicator measurement result.

[0281] In another possible implementation, the filtered indicator measurement result includes a model indicator and / or system performance of the first model; the inference management includes: if the model indicator and / or system performance is less than or equal to a fourth threshold value, falling back to a default mode, the default mode being to perform transmission in a non-AI manner, or measuring an indicator of a third model, or switching to a fourth model; or if the model indicator and / or system performance is greater than or equal to a fifth threshold value, activating the first model or switching to the first model; or if the model indicator and / or system performance of the first model measured by the first device 800 for Q consecutive times is less than or equal to a sixth threshold value, requesting to train the first model or requesting to collect a data set, the data set being used to train the first model, Q being an integer greater than or equal to 1.

[0282] In another possible implementation, the filtered indicator measurement result includes a model indicator and / or system performance of the first model; the training management includes: if the model indicator and / or system performance is less than or equal to a seventh threshold value, terminating the training of the first model or adjusting a learning rate of the first model; or if the model indicator and / or system performance is greater than or equal to an eighth threshold value, initializing the first model or loading a second model, the first model being obtained by training the AI service based on the first data for M rounds, the second model being a model obtained by training the AI service for M-1 rounds, M being an integer greater than or equal to 2.

[0283] It should be understood that the specific processes in which each module performs the corresponding processes described above have been described in detail in the method embodiments described above, and thus are not described herein again for the sake of brevity.

[0284] The processing module 802 in the above embodiments can be implemented by at least one processor or processor-related circuit. The transceiver module 801 can be implemented by a transceiver or transceiver-related circuit. The transceiver module 801 can also be referred to as a communication module or a communication interface. The storage module can be implemented by at least one memory.

[0285] FIG. 9 is a structural schematic diagram of a second device according to an embodiment of the present application. Referring to FIG. 9, the second device can be used to execute the process executed by the second device in the embodiment shown in FIG. 3. For details, please refer to the related description in the above method embodiments.

[0286] The second device 900 includes a transceiver module 901 and a processing module 902.

[0287] The processing module 902 is configured to perform data processing. The transceiver module 901 can implement corresponding communication functions. The transceiver module 901 can also be referred to as a communication interface or a communication module.

[0288] Optionally, the second device 900 can further include a storage module, which can be used to store instructions and / or data. The processing module 902 can read the instructions and / or data in the storage module, so that the second device implements the above method embodiments.

[0289] The second device 900 can be used to execute the actions performed by the second device in the above method embodiments. The second device 900 can be the second device or a component configurable to the second device. The processing module 902 is configured to perform the processing-related operations of the second device side in the above method embodiments. The transceiver module 901 is configured to perform the receiving-related operations of the second device side in the above method embodiments.

[0290] Optionally, the transceiver module 901 can include a sending module and a receiving module. The sending module is configured to perform the sending operations in the above method embodiments. The receiving module is configured to perform the receiving operations in the above method embodiments.

[0291] It should be noted that the second device 900 can include a sending module but not a receiving module. Alternatively, the second device 900 can include a receiving module but not a sending module. Whether the second device 900 includes a sending module or a receiving module can depend on whether the second device 900 performs the sending actions and the receiving actions in the above schemes.

[0292] Optionally, the second device 900 is configured to perform the actions performed by the second device in the above embodiment shown in FIG. 3. For details, please refer to the related description in the above embodiment shown in FIG. 3, which will not be described here in detail. For example, the second device 900 is configured to perform the following scheme:

[0293] The transceiver module 901 is configured to receive the filtered index measurement result from the first device, the filtered index measurement result being obtained by filtering the first index measurement result, the first index measurement result being determined according to the first data and the index of the first model;

[0294] The processing module 902 is configured to manage the first model according to the filtered index measurement result.

[0295] For other implementation manners, refer to the detailed description of the embodiment shown in FIG. 3.

[0296] It should be understood that the specific processes in which the modules perform the corresponding processes are described in the foregoing method embodiments, and thus are not described herein again for the sake of brevity.

[0297] The processing module 902 in the foregoing embodiments can be implemented by at least one processor or processor-related circuit. The transceiver module 901 can be implemented by a transceiver or transceiver-related circuit. The transceiver module 901 can also be referred to as a communication module or a communication interface. The storage module can be implemented by at least one memory.

[0298] FIG. 10 is a structural schematic diagram of a second device according to an embodiment of the present application. Referring to FIG. 10, the second device can be configured to perform the processes performed by the second device in the embodiment shown in FIG. 3. For details, refer to the related description in the foregoing method embodiments.

[0299] The second device 1000 includes a transceiver module 1001. Optionally, the second device 1000 further includes a processing module 1002.

[0300] The processing module 1002 is configured to perform data processing. The transceiver module 1001 can implement corresponding communication functions. The transceiver module 1001 can also be referred to as a communication interface or a communication module.

[0301] Optionally, the second device 1000 can further include a storage module, which can be configured to store instructions and / or data. The processing module 1002 can read the instructions and / or data in the storage module, so that the second device implements the foregoing method embodiments.

[0302] The second device 1000 can be configured to perform the actions performed by the second device in the foregoing method embodiments. The second device 1000 can be the second device or a component configurable to the second device. The processing module 1002 is configured to perform the processing-related operations of the second device side in the foregoing method embodiments. The transceiver module 1001 is configured to perform the receiving-related operations of the second device side in the foregoing method embodiments.

[0303] Optionally, the transceiver module 1001 can include a sending module and a receiving module. The sending module is configured to perform the sending operations in the above method embodiments. The receiving module is configured to perform the receiving operations in the above method embodiments.

[0304] It should be noted that the second device 1000 can include a sending module, but not a receiving module. Alternatively, the second device 1000 can include a receiving module, but not a sending module. Specifically, whether the second device 1000 includes a sending action and a receiving action in the above scheme can be determined.

[0305] Optionally, the second device 1000 is configured to perform the actions performed by the second device in the above embodiment shown in FIG. 3. For details, refer to the related description in the above embodiment shown in FIG. 3, which will not be repeated here. For example, the second device 1000 is configured to perform the following scheme:

[0306] The transceiver module 1001 is configured to receive the management result of the first model from the first device. The management result is obtained by managing the first model according to the filtered index measurement result. The filtered index measurement result is obtained by filtering the first index measurement result. The first index measurement result is determined by measuring the index of the first model according to the first data.

[0307] For other implementation manners, refer to the detailed description of the above embodiment shown in FIG. 3, which will not be repeated here.

[0308] It should be understood that the specific processes of each module performing the above corresponding processes have been described in detail in the above method embodiments. For brevity, they will not be repeated here.

[0309] The processing module 1002 in the above embodiment can be implemented by at least one processor or processor-related circuit. The transceiver module 1001 can be implemented by a transceiver or transceiver-related circuit. The transceiver module 1001 can also be referred to as a communication module or a communication interface. The storage module can be implemented by at least one memory.

[0310] The embodiment of the present application also provides a device 1100. Please refer to FIG. 11. The device 1100 includes a processor 1110 and a memory 1120. The memory 1120 is configured to store computer programs or instructions and / or data. The processor 1110 is configured to execute the computer programs or instructions and / or data stored in the memory 1120, so that the method in the above method embodiments is performed. The device 1100 is configured to implement the operations performed by the first device or the second device in the above method embodiments.

[0311] Optionally, the processor 1110 included in the device 1100 is one or more.

[0312] Optionally, the apparatus 1100 can further include a memory 1120, as shown in FIG. 11.

[0313] Optionally, the memory 1120 included in the apparatus 1100 can be one or more.

[0314] Optionally, the memory 1120 can be integrated with the processor 1110, or be separately arranged.

[0315] Optionally, the apparatus 1100 can further include a transceiver 1130 for receiving and / or sending signals, as shown in FIG. 11. For example, the processor 1110 is configured to control the transceiver 1130 to receive and / or send signals.

[0316] The present application also provides an apparatus 1200, which can be a terminal device, a processor in a terminal device, or a chip. The apparatus 1200 can be configured to perform operations performed by the first apparatus in the above method embodiments.

[0317] When the apparatus 1200 is a terminal device, FIG. 12 shows a simplified structural schematic diagram of the terminal device. As shown in FIG. 12, the terminal device includes a processor, a memory, and a transceiver. The memory can store computer program codes, and the transceiver includes a transmitter 1231, a receiver 1232, a radio frequency circuit (not shown in the figure), an antenna 1233, and an input / output device (not shown in the figure).

[0318] The processor is mainly configured to process communication protocols and communication data, control the terminal device, execute software programs and process data of the software programs, and the like.

[0319] The memory is mainly configured to store software programs and data.

[0320] The radio frequency circuit is mainly configured to convert baseband signals and radio frequency signals, and process radio frequency signals.

[0321] The antenna is mainly configured to transceive radio frequency signals in the form of electromagnetic waves.

[0322] The input / output device can include a touch screen, a display screen, a keyboard, or the like. The input / output device is mainly configured to receive data input by a user and output data to the user. It should be noted that some types of terminal devices can not have an input / output device.

[0323] When data needs to be sent, the processor performs baseband processing on the data to be sent, and outputs a baseband signal to the radio frequency circuit. Then, the radio frequency circuit performs radio frequency processing on the baseband signal, and sends a radio frequency signal in the form of an electromagnetic wave through an antenna. When data is sent to the terminal device, the radio frequency circuit receives a radio frequency signal through the antenna. The radio frequency circuit converts the radio frequency signal into a baseband signal, and outputs the baseband signal to the processor. The processor converts the baseband signal into data and processes the data. For ease of illustration, only one memory, one processor, and one transceiver are shown in FIG. 12. In an actual terminal device product, there can be one or more processors and one or more memories. The memory can also be referred to as a storage medium or a storage device, etc. The memory can be independent of the processor, or can be integrated with the processor, and the embodiments of the present application do not limit this.

[0324] In the embodiments of the present application, the antenna and the radio frequency circuit with transceiving functions can be regarded as a transceiving module of the terminal device, and the processor with processing functions can be regarded as a processing module of the terminal device.

[0325] As shown in FIG. 12, the terminal device includes a processor 1210, a memory 1220, and a transceiver 1230. The processor 1210 can also be referred to as a processing unit, a processing board, a processing module, or a processing device, etc. The transceiver 1230 can also be referred to as a transceiving unit, a transceiver, or a transceiving device, etc.

[0326] Optionally, the devices for implementing the receiving function in the transceiver 1230 are regarded as a receiving module, and the devices for implementing the sending function in the transceiver 1230 are regarded as a sending module, that is, the transceiver 1230 includes a receiver and a transmitter. The transceiver can also be referred to as a transceiver, a transceiving module, or a transceiving circuit, etc. The receiver can also be referred to as a receiver, a receiving module, or a receiving circuit, etc. The transmitter can also be referred to as a transmitter, a transmitting module, or a transmitting circuit, etc.

[0327] The processor 1210 is configured to perform the processing actions of the first device side in the embodiments shown in FIG. 3. The transceiver 1230 is configured to perform the transceiving actions of the first device side in the embodiments shown in FIG. 3.

[0328] It should be understood that FIG. 12 is merely an example and not limiting, and the above terminal device including a transceiving module and a processing module can not depend on the structure shown in FIG. 8 or FIG. 12.

[0329] When the apparatus 1200 is a chip, the chip includes a processor, a memory and a transceiver. The transceiver can be an input output circuit or a communication interface. The processor can be an integrated processing module or a microprocessor or an integrated circuit. The sending operation of the first apparatus in the above method embodiments can be understood as the output of the chip, and the receiving operation of the first apparatus in the above method embodiments can be understood as the input of the chip.

[0330] The application further provides an apparatus 1300, which can be a network device or a chip. The apparatus 1300 can be used to perform the operations performed by the second apparatus in the embodiment shown in FIG. 3.

[0331] When the apparatus 1300 is a network device, for example, a base station. FIG. 13 shows a simplified structure diagram of a base station. The base station includes a 1310 part, a 1320 part and a 1330 part.

[0332] The 1310 part is mainly used for baseband processing, controlling the base station and the like; the 1310 part is usually the control center of the base station, which can be usually referred to as a processor, and is used to control the base station to perform the processing operations of the second apparatus in the above method embodiments.

[0333] The 1320 part is mainly used for storing computer program codes and data.

[0334] The 1330 part is mainly used for transceiving radio frequency signals and converting radio frequency signals and baseband signals; the 1330 part can be usually referred to as a transceiving module, a transceiver, a transceiving circuit or a transceiver and the like. The transceiving module of the 1330 part can also be referred to as a transceiver or a transceiver and the like, which includes an antenna 1333 and a radio frequency circuit (not shown in the figure), wherein the radio frequency circuit is mainly used for radio frequency processing. Optionally, the devices for realizing the receiving function in the 1330 part can be regarded as a receiver, and the devices for realizing the sending function can be regarded as a transmitter, that is, the 1330 part includes a receiver 1332 and a transmitter 1331. The receiver can also be referred to as a receiving module, a receiver or a receiving circuit and the like, and the transmitter can be referred to as a transmitting module, a transmitter or a transmitting circuit and the like.

[0335] The 1310 part and the 1320 part can include one or more single boards, and each single board can include one or more processors and one or more memories. The processor is used to read and execute the program in the memory to realize the baseband processing function and the control of the base station. If there are multiple single boards, the single boards can be interconnected to enhance the processing capability. As an optional implementation, the multiple single boards can share one or more processors, or the multiple single boards can share one or more memories, or the multiple single boards can share one or more processors at the same time.

[0336] For example, in an implementation, the transceiver module of the 1330 part is configured to perform the transceiving related procedures performed by the second device in the embodiments shown in FIG. 3. The processor of the 1310 part is configured to perform the processing related procedures performed by the second device in the embodiments shown in FIG. 3.

[0337] It should be understood that FIG. 13 is merely an example but not a limitation, and the network device including the processor, the memory and the transceiver described above can not depend on the structure shown in FIG. 9, FIG. 10 or FIG. 13.

[0338] When the device 1300 is a chip, the chip includes a transceiver, a memory and a processor. The transceiver can be an input / output circuit, a communication interface; the processor is a processor integrated on the chip, or a microprocessor, or an integrated circuit. The transmitting operation of the second device in the method embodiments described above can be understood as the output of the chip, and the receiving operation of the second device in the method embodiments described above can be understood as the input of the chip.

[0339] The embodiments of the present application further provide a computer readable storage medium, having stored thereon computer instructions for implementing the method performed by the first device or the second device in the method embodiments described above.

[0340] For example, the computer program is executed by a computer, so that the computer can implement the method performed by the first device or the second device in the method embodiments described above.

[0341] The embodiments of the present application further provide a computer program product including instructions, which are executed by a computer to make the computer implement the method performed by the first device or the second device in the method embodiments described above.

[0342] The embodiments of the present application further provide a communication system, including the first device in the embodiments above and the second device in the embodiments above. The first device is configured to perform part or all of the operations performed by the first device in the method embodiments above, and the second device is configured to perform part or all of the operations performed by the second device in the method embodiments above.

[0343] The embodiments of the present application further provide a chip device including a processor, which is configured to invoke computer degrees or computer instructions stored in the memory, so that the processor performs the method provided by the embodiments shown in FIG. 3.

[0344] In a possible implementation, the input of the chip device corresponds to the receiving operation in any one of the embodiments shown in FIG. 3, and the output of the chip device corresponds to the transmitting operation in any one of the embodiments shown in FIG. 3.

[0345] Optionally, the processor is coupled with the memory through an interface.

[0346] Optionally, the chip device further comprises a memory, and the memory stores computer programs or computer instructions.

[0347] The processor mentioned in any of the above can be a general central processing unit, a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of programs for providing the method of any of the embodiments shown in Fig. 3. The memory mentioned in any of the above can be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM), and the like.

[0348] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the explanation and beneficial effects of the related content in any of the above-mentioned devices can refer to the corresponding method embodiments provided above, which will not be repeated here.

[0349] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the above-mentioned device embodiments are only schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0350] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or they can be distributed on a plurality of network units. According to actual needs, some or all of the units can be selected to achieve the purpose of the embodiment.

[0351] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0352] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the part of the technical solutions of the present application that essentially makes contributions or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of 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 methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a ROM, a RAM, a magnetic disk, or an optical disk, and various media that can store program codes.

[0353] The above-described embodiments are merely used to illustrate the technical solutions of the present application, rather than limit the technical solutions thereof; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions recorded in the foregoing embodiments can still be modified, or some technical features thereof can be replaced by equivalents; and such modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A communication method characterized by comprising: The method comprises: The first device determines a first index measurement result according to a first data measurement of an index of a first model; The first device filters the first index measurement result to obtain a filtered index measurement result; The first device sends the filtered index measurement result to a second device, or the first device manages the first model according to the filtered index measurement result and sends a management result of the first model to the second device.

2. A communication method characterized by comprising: The method comprises: The second device receives a filtered index measurement result from a first device, the filtered index measurement result being obtained by filtering a first index measurement result, the first index measurement result being determined according to a first data measurement of an index of a first model; The second device manages the first model according to the filtered index measurement result.

3. A communication method characterized by comprising: The method comprises: The second device receives a management result of a first model from a first device, the management result being obtained by managing the first model according to a filtered index measurement result, the filtered index measurement result being obtained by filtering a first index measurement result, the first index measurement result being determined according to a first data measurement of an index of the first model.

4. The method of claim 1, wherein, The method further comprises: The first device receives a management confirmation from the second device, the management confirmation being used to determine the management result.

5. The method of claim 3, wherein, The method further comprises: The second device sends a management confirmation to the first device, the management confirmation being used to determine the management result.

6. The method according to claim 1 or 4, characterized in that, The method further comprises: The first device receives first configuration information from the second device, the first configuration information being used to indicate a target filtering manner; The first device filters the first index measurement result to obtain a filtered index measurement result, comprising: The first device filters the first index measurement result according to the target filtering manner to obtain the filtered index measurement result.

7. The method of claim 2, 3, or 5, wherein, The method further comprises: The second device sends first configuration information to the first device, the first configuration information being used to indicate a target filtering manner, the target filtering manner being used to filter the first index measurement result.

8. The method according to claim 6 or 7, characterized in that, The target filtering manner is a first filtering manner, the first filtering manner comprising weighting processing the first index measurement result and a historical filtered index measurement result, the historical filtered index measurement result comprising a filtered index measurement result obtained by filtering an index of the first model measured by the first device before the first index measurement result; or, The target filtering manner is a second filtering manner, the second filtering manner comprising weighting processing the first index measurement result and a historical filtered index measurement result, the historical filtered index measurement result comprising a filtered index measurement result obtained by filtering an index of the first model measured by the first device before the first index measurement result. The target filtering manner is a second filtering manner, and the second filtering manner comprises weighting the first index measurement result and a historical index measurement result, the first index measurement result being obtained by the first device measuring an index of the first model at an Nth time or an Nth time, and the historical index measurement result comprising an index measurement result obtained by the first device measuring the index of the first model at K previous times of the Nth time or K previous times of the Nth time, K being an integer greater than or equal to 1, and N being an integer greater than or equal to 2.

9. The method according to any one of claims 6 to 8, characterized in that, The first configuration information comprises a first weighting factor used for weighting the first index measurement result and the historical filtered index measurement result, or the first configuration information comprises at least one second weighting factor used for weighting the first index measurement result and the historical index measurement result and / or the K.

10. The method of any one of claims 1, 4, 6, 8, 9, wherein, The method further comprises: The first device receives second configuration information from the second device, the second configuration information being used for indicating a measurement configuration. The first device measures an index of a first model according to first data to determine a first index measurement result, comprising: The first device measures the index of the first model according to the first data and the measurement configuration to determine the first index measurement result.

11. The method according to any one of claims 2, 3, 5, 7 to 9, characterized in that, The method further comprises: The second device sends second configuration information to the first device, the second configuration information being used for indicating a measurement configuration, and the measurement configuration being used for the first device to measure the index of the first model.

12. The method according to claim 10 or 11, characterized in that, The measurement configuration comprises at least one of the following: an index to be measured or a number of samples used for determining the index to be measured.

13. The method of claim 12, wherein, The index to be measured comprises at least one of the following: a model index or system performance, and the model index comprises at least one of the following: accuracy of a model inference result, loss function value of model training, variation of a model parameter of model training, performance of a validation data set, or input-output distribution drift.

14. The method of claims 1, 4, 6, 8-10, 12, 13, wherein, The first device measures an index of a first model according to first data to determine a first index measurement result, comprising: The first device measures the index of the first model according to the first data to obtain a second index measurement result. The first device performs anomaly detection on the first data according to the second index measurement result to obtain abnormal data. The first device determines a first index measurement result, the first index measurement result being an index measurement result other than an index measurement result corresponding to the abnormal data in the second index measurement result.

15. The method of claim 14, wherein, The first data comprises one or more data, and the abnormal data comprises data whose value is greater than or equal to a first threshold value with respect to a reference data value.

16. The method according to claim 14 or 15, characterized in that The second index measurement result includes one or more model indexes, and the index measurement result corresponding to the abnormal data is an index measurement result whose difference from a reference model index is greater than or equal to a second threshold value among the one or more model indexes; or The second index measurement result includes one or more system performances, and the index measurement result corresponding to the abnormal data is an index measurement result whose difference from a reference system performance is greater than or equal to a third threshold value among the one or more system performances.

17. The method according to any one of claims 14 to 16, characterized in that, The method further includes: The first device sends an abnormal result to the second device, the abnormal result including the abnormal data and / or a proportion of the abnormal data.

18. The method according to any one of claims 2, 3, 5, 7 to 9, 11 to 13, characterized in that, The method further includes: The second device receives an abnormal result from the first device, the abnormal result including the abnormal data and / or a proportion of the abnormal data.

19. The method of claim 14, 15, or 17, wherein, The first device manages the first model according to the filtered index measurement result, including: The first device loads a second model, the first model being obtained by M rounds of training based on the first data for an artificial intelligence (AI) service, the second model being a model obtained by M-1 rounds of training for the AI service, M being an integer greater than or equal to 2; or The first device updates a model parameter of the first model.

20. The method of any one of claims 1, 4, 6-10, 12-17, 19, wherein, The method further includes: The first device receives third configuration information from the second device, the third configuration information being used to indicate a reporting configuration; The first device sends the filtered index measurement result to the second device, including: The first device sends the filtered index measurement result to the second device according to the reporting configuration.

21. The method of any one of claims 2, 3, 5, 7-9, 11-13, 18, wherein, The method further includes: The second device sends third configuration information to the first device, the third configuration information being used to indicate a reporting configuration, the reporting configuration being used to report the filtered index measurement result.

22. The method of claim 21, wherein, The reporting configuration includes at least one of a reporting condition or a reporting period.

23. The method of any one of claims 1, 4, 6, 8-10, 12-17, 19, 20, 22, wherein, The first device manages the first model according to the filtered index measurement result, including: The first device performs inference management or training management on the first model according to the filtered index measurement result.

24. The method of any one of claims 2, 7-9, 11-13, 17, 18, 21, 22, wherein, The second device manages the first model according to the filtered index measurement result, including: The second device performs inference management or training management on the first model according to the filtered index measurement result.

25. The method of claim 23 or 24, wherein, The filtered index measurement result includes a model index and / or a system performance of the first model; and the inference management includes: If the model index and / or the system performance is less than or equal to a fourth threshold value, falling back to a default mode, the default mode being transmission in a non-artificial intelligence (AI) manner, or measuring an index of a third model, or switching to a fourth model; Or If the model index and / or the system performance is greater than or equal to a fifth threshold value, activating the first model, or switching to the first model; Or If the model index and / or the system performance of the first model measured by the first device for Q consecutive times is less than or equal to a sixth threshold value, the first model is requested to be trained, or a data set used for training the first model is requested to be collected, where Q is an integer greater than or equal to 1.

26. The method of claim 23 or 24, wherein, The filtered index measurement result includes the model index and / or the system performance of the first model; the training management includes: If the model index and / or the system performance is less than or equal to a seventh threshold value, the training of the first model is terminated, or the learning rate of the first model is adjusted; or, If the model index and / or the system performance is greater than or equal to an eighth threshold value, the first model is initialized or a second model is loaded, the first model is obtained by training the first data for M rounds of training of an artificial intelligence (AI) service, and the second model is a model obtained by training the AI service for M-1 rounds of training, where M is an integer greater than or equal to 2.

27. An apparatus comprising: The device includes a transceiver module and a processing module; The transceiver module is configured to perform the transceiving operation of the method in any one of claims 1, 4, 6, 8-10, 12-16, 17, 19, 20, 22, 23, 25, and 26; and the processing module is configured to perform the processing operation of the method in any one of claims 1, 4, 6, 8-10, 12-16, 17, 19, 20, 22, 23, 25, and 26; or The transceiver module is configured to perform the transceiving operation of the method in any one of claims 2, 7-9, 11-13, 18, 21, 22, 24-26; and the processing module is configured to perform the processing operation of the method in any one of claims 2, 7-9, 11-13, 18, 21, 22, 24-26.

28. An apparatus comprising: The device includes a transceiver module configured to perform the transceiving operation of the method in any one of claims 3, 5, 7-9, 11-13, 18, 21, and 22.

29. The apparatus of claim 28, wherein, The device further includes a processing module configured to perform the processing operation of the method in any one of claims 3, 5, 7-9, 11-13, 18, 21, and 22.

30. An apparatus comprising: The device includes a processor configured to execute computer programs or computer instructions in a memory to perform the method in any one of claims 1-26.

31. A computer readable storage medium, characterized in that, A computer program is stored on the device, and when the computer program is executed by the device, the device performs the method in any one of claims 1-26.

32. A computer program product, characterised in that, The computer program product includes computer instructions, which, when the computer program product is run on a computer, cause the computer to perform the method in any one of claims 1-26.

Images