CN-122001423-A - Communication method and device

Abstract

A communication method and apparatus are provided. In the method, the network side not only can send a data set and/or a model for model training, but also can indicate a plurality of groups of indexes corresponding to the data set and/or the model, wherein the plurality of groups of indexes correspond to an reasoning task of the first model, and the reasoning task can be to compress target CSI. The terminal side may perform model training based on one or more of the multiple sets of metrics and the data set and/or the model to obtain a first model. The terminal side can perform model training based on one or more groups of corresponding indexes after acquiring the data set and/or the model. Thereby being beneficial to the terminal side to obtain a model meeting the requirements through model training, and further improving the performance of CSI feedback.

Inventors

• TIAN YANG
• LI YUAN
• SUN YAN

Assignees

• 华为技术有限公司

Dates

Publication Date

20260508

Application Date

20241108

Claims (20)

1. 1. A method of communication, comprising: acquiring a data set and/or a model, wherein the data set and/or the model are used for model training; And training the model based on a first index and the data set and/or the model to obtain a first model, wherein the first index is one or more of multiple groups of indexes, the multiple groups of indexes comprise one or more of performance indexes, indexes of time length required for training or indexes of resources required for training, the multiple groups of indexes correspond to the data set and/or the model and correspond to an reasoning task of the first model, and the reasoning task of the first model is to compress target Channel State Information (CSI).
2. 2. The method of claim 1, wherein prior to the model training based on the first metric, and the dataset and/or the model, the method further comprises: first information is received, the first information being used to indicate the plurality of sets of metrics.
3. 3. The method of claim 2, wherein the method further comprises: and acquiring the multiple groups of indexes according to the first information.
4. 4. A method according to any one of claims 1 to 3, wherein the first indicator is the performance indicator, at least two of the plurality of sets of indicators are the performance indicator, and the first indicator is one or more of the at least two sets of indicators.
5. 5. The method of claim 4, wherein the obtaining the data set and/or model comprises obtaining a first data set, the first data set comprising target CSI from a network side and CSI feedback information, the CSI feedback information being obtained by compressing the target CSI.
6. 6. The method of claim 5, wherein the first data set is used to train a second model, the second model being used to train the first model, wherein an inference task of the second model is to decompress the CSI feedback information; the at least two sets of functions corresponding to the at least two sets of indicators comprise a first set of functions comprising one or more functions selected from a normalized mean square error NMSE, mean square error MSE, L1Loss, generalized cosine similarity GCS, square generalized cosine similarity SGCS between the output of the second model and a label, or a weighted sum of a plurality of NMSE, MSE, L Loss, GCS or SGCS between the output of the second model and the label.
7. 7. The method of claim 5, wherein the first data set is used to train the first model; at least two sets of functions corresponding to the at least two sets of metrics include a second set of functions including one or more of NMSE, MSE, L a Loss, GCS, SGCS between the output of the first model and the CSI feedback information, or a weighted sum of a plurality of NMSE, MSE, L a Loss, GCS, or SGCS between the output of the first model and the CSI feedback information.
8. 8. The method of claim 4, wherein the acquiring the dataset and/or the model comprises acquiring a third model whose reasoning task is to compress the target CSI.
9. 9. The method of claim 8, wherein the third model is used to train a second model, the second model is used to train the first model, wherein an inference task of the second model is to decompress CSI feedback information obtained by compressing the target CSI by the third model; The at least two sets of functions corresponding to the at least two sets of metrics include a third set of functions comprising one or more functions of NMSE, MSE, L a Loss, GCS, SGCS between the output of the second model and the tag, or a weighted sum of a plurality of NMSE, MSE, L a Loss, GCS, or SGCS between the output of the second model and the tag.
10. 10. The method of claim 8, wherein the third model is used to obtain a second data set comprising CSI feedback information, the second data set being used to train the first model; The at least two sets of functions corresponding to the at least two sets of metrics include a fourth set of functions comprising one or more of NMSE, MSE, L los between the output of the first model and the CSI feedback information or a weighted sum of the multiple of NMSE, MSE, or L1Loss between the output of the first model and the CSI feedback information.
11. 11. The method of claim 6 or 9, wherein the outputting of the second model comprises: the output of the second model when training the second model, and/or And when training the first model according to the second model, outputting the second model.
12. 12. The method of any one of claims 1 to 11, wherein the method further comprises: and executing the reasoning task of the first model under the condition that the first index is met.
13. 13. The method of any one of claims 1 to 12, wherein the method further comprises: and sending second information, wherein the second information is used for indicating that the first index is met.
14. 14. The method of any one of claims 1 to 13, wherein the method further comprises: Transmitting third information for one or more of the following if the first index is not satisfied: Indicating that the first indicator is not satisfied; Requesting replacement of the dataset; Requesting replacement of the model, or Requesting to stop or pause the model training.
15. 15. The method according to any one of claims 1 to 14, wherein the acquiring a dataset and/or a model comprises: The data set and/or the model are received.
16. 16. A method of communication, comprising: Transmitting a data set and/or a model, the data set and/or the model being used for model training; The method comprises the steps of sending first information, wherein the first information is used for indicating a plurality of groups of indexes, the plurality of groups of indexes are used for model training, the plurality of groups of indexes comprise one or more indexes of performance indexes, indexes of time length required for training or indexes of resources required for training, the plurality of groups of indexes correspond to a data set and/or a model and correspond to an reasoning task of the first model, and the reasoning task of the first model is used for compressing target Channel State Information (CSI).
17. 17. The method of claim 16, wherein at least two of the plurality of sets of metrics are the performance metrics.
18. 18. The method of claim 17, wherein transmitting the data set and/or model comprises transmitting a first data set comprising target CSI and CSI feedback information, the CSI feedback information being compressed from the target CSI.
19. 19. The method of claim 18, wherein the first data set is used to train a second model, the second model being used to train the first model, wherein an inference task of the second model is to decompress the CSI feedback information; the at least two sets of functions corresponding to the at least two sets of indicators comprise a first set of functions comprising one or more functions selected from a normalized mean square error NMSE, mean square error MSE, L1Loss, generalized cosine similarity GCS, square generalized cosine similarity SGCS between the output of the second model and a label, or a weighted sum of a plurality of NMSE, MSE, L Loss, GCS or SGCS between the output of the second model and the label.
20. 20. The method of claim 18, wherein the first data set is used to train the first model; at least two sets of functions corresponding to the at least two sets of metrics include a second set of functions including one or more of NMSE, MSE, L a Loss, GCS, SGCS between the output of the first model and the CSI feedback information, or a weighted sum of a plurality of NMSE, MSE, L a Loss, GCS, or SGCS between the output of the first model and the CSI feedback information.

Description

Communication method and device Technical Field The present application relates to the field of wireless communications, and in particular, to a communication method and apparatus. Background In a wireless communication system, an artificial intelligence (ARTIFICIAL INTELLIGENCE, AI) model may be employed to compress and reconstruct (or otherwise recover) Channel State Information (CSI). The sender (such as terminal equipment) of the CSI report can compress CSI through an AI model, and send out the obtained CSI report. The receiver of the CSI report (e.g., a network device) may reconstruct the channel measurement results based on the received CSI report through the AI model. In order to support the interfacing and model development of double-ended AI models, such as models deployed at the sender and receiver of CSI reports, a data set (dataset) and/or model needs to be provided for the sender and/or receiver for model training. One possible implementation is that the sender provides the receiver with a data set and/or a model for model training of the receiver. However, the model training of the receiver may not match the data set or model provided by the sender, resulting in an unsatisfactory or even unusable model. Disclosure of Invention The application provides a communication method and a communication device, which are beneficial to training a model meeting requirements and improving the feedback performance of CSI. In a first aspect, a communication method is provided, which may be applied to a terminal side, for example, may be performed by a terminal device, or may be performed by a component disposed in the terminal device, for example, a circuit or a chip inside the terminal device (such as a modem (modem) chip, also called a baseband (baseband) chip, or a system on chip (SoC) chip or a system in package (SYSTEM IN PACKAGE) chip including a modem core, or the like), or may be performed by a device disposed outside the terminal device (such as a host of an Over The Top (OTT) system or a cloud server) or a component in the device (such as a chip, a processor, or a circuit inside the device), or may be implemented by a logic module or software capable of implementing all or part of the functions of the terminal device, or the like. Alternatively, the method may be performed by a first apparatus, which may be a terminal device, or may be a component in the terminal device, such as a circuit or a chip inside the terminal device (such as a modem chip), or a SoC chip or a SIP chip containing the modem core, or may be a device outside the terminal device (such as a host of an OTT system or a cloud server) or a component in the device (such as a chip inside the device, a processor, or a circuit, or the like), or may be a logic module or software capable of implementing part or all of functions on the terminal side, or the like. The application is not limited in this regard. For ease of understanding and explanation, the method provided in the first aspect is described below by taking the terminal side as an example. The method comprises the steps of obtaining a data set and/or a model, wherein the data set and/or the model are used for model training, conducting model training based on a first index and the data set and/or the model to obtain a first model, wherein the first index is one or more of multiple groups of indexes, the multiple groups of indexes comprise one or more of performance indexes, indexes of time length required for training or indexes of resources required for training, the multiple groups of indexes correspond to the data set and/or the model, and correspond to reasoning tasks of the first model, and the reasoning tasks of the first model are compression of target CSI. The first model may be obtained by performing model training on the terminal device, or may be obtained by training other devices except the terminal device, which is not limited in the present application. The data set may be sent from the network side to the terminal side. The acquiring of the data set and/or model by the terminal side may include that the terminal device of the terminal side may receive the data set and/or model from the network device of the network side, or that the host or cloud server of the OTT system of the terminal side may receive the data set and/or model from the intelligent network element of the network side, or that the terminal device of the terminal side may acquire the data set and/or model received from the intelligent network element from the host or cloud service area of the OTT system, or that the host or cloud server of the OTT system of the terminal side may acquire the data set and/or model received from the network device from the terminal device. Based on the above-described scheme, the first index corresponds not only to the data set and/or model used for model training, but also to the reasoning task of the first model. The terminal side can acquire the data set and/or the