A management method and related device for machine learning tasks

Abstract

This application discloses a machine learning task management method and related devices, which can be applied in the process of federated learning. Obtain the machine learning task sent by the coordination device; then call the training data based on the machine learning task to train the first machine learning model to obtain the second machine learning model; and test the second machine learning model in a trusted execution environment ; If the verification of the second machine learning model is passed, sending the second machine learning model to the coordinating device. Since the training results are verified in the trusted execution environment, the authenticity of the workload proof of the participating devices is guaranteed, and the participating devices that do not normally execute the machine learning task are excluded, thereby improving the accuracy of the training results in the machine learning task.

Description

technical field [0001] The present application relates to the field of computer technology, and in particular to a method for managing machine learning tasks and related devices. Background technique [0002] In today's society, machine learning is increasingly integrated into people's lives, such as product recommendation, facial recognition payment, and game AI. Machine learning is to obtain knowledge from data, and whether the data is sufficient or not largely affects the effect of machine learning. Data related to people's lives often involves user privacy, such as chat records, disease inspection results, and photos. While enjoying the convenience brought by machine learning, people also expect their data to be effectively protected. [0003] In general, machine learning algorithms usually concentrate data on servers for training, which threatens people's privacy to a certain extent. In the recent federated learning, training is performed at the data owner, and only ...

AI Technical Summary

Problems solved by technology

[0004] However, since some participating devices may perform the training process of federated learning by forging proof of work, it will affect the accuracy of the training results of machine learning models similar to federated learning

Images