Method, system and device for model training

Abstract

This description relates to a model training method, system and device, which can be used for privacy data protection in the process of machine learning. The method is implemented by any training member of the participating parties, including: performing model training based on the training samples held by oneself Training to obtain the model data list; select several sub-matrices from the sub-matrices corresponding to one or more convolution kernels in the model data list, and determine the first position information of the selected sub-matrix in the model data list; from the model data Select some non-convolution kernel elements in the list except the sub-matrix corresponding to the convolution kernel, and determine to reflect the second position information of the selected non-convolution kernel elements in the model data list; determine to include the first position information and The transmission data list of the second location information is uploaded to the server; the model data update list issued by the server is obtained, and the local model is updated based on the model data update list.

Description

technical field [0001] One or more embodiments of this specification relate to multi-party collaborative model training, and in particular to a method, system and device for model training. Background technique [0002] In the fields of data analysis, data mining, economic forecasting, etc., federated learning can collaboratively train machine learning models for multiple parties to use while ensuring the security of multi-party data. However, when federated learning is applied to a model with a large amount of calculation and many parameters, the model parameters that need to be transmitted will also increase accordingly, which will cause a large amount of data transmission during the federated learning process, so the pressure on communication transmission will be greater. [0003] Therefore, it is necessary to propose a model training method to reduce the transmission pressure in the federated learning process. Contents of the invention [0004] One aspect of this spec...

AI Technical Summary

Problems solved by technology

However, when federated learning is applied to the situation where the model has a large amount of calculation and many parameters, the model parameters that need to be transmitted will also increase accordingly, which will cause a large amount of data transmission during the federated learning process, so the pressure on communication transmission is greater

Images