Method, device and alliance learning system for alliance learning

Abstract

Embodiments of the present specification provide methods for federated learning, the federation including at least three first member devices and second member devices, each first member device having local data. In this method, the second member device groups the first member devices based on the hardware performance parameter information of each first member device to obtain each first member device group, and each first member device group has the same hardware The initial global business model for performance parameter adaptation. For each first member device group, each first member device in the first member device group uses respective local data to perform federation learning together with the second member device to obtain the target global business model.

Description

technical field [0001] The embodiments of this specification generally relate to the field of artificial intelligence, and in particular, relate to a method, an apparatus, and a federated learning system for federated learning. Background technique [0002] With the development of artificial intelligence technology, business models have been gradually applied to various business application scenarios, such as risk assessment, speech recognition, natural language processing, etc. In order to solve the problem of data islands during model training, Federated Learning is proposed. Using federated learning, it is possible to allow multiple data owners to perform encrypted calculations through multi-party secure computing or in a hardware SGX environment while protecting data privacy, and to achieve model training through collaborative computing among various data owners. [0003] However, in the federated learning scenario, each data owner's device may be a terminal device with...

AI Technical Summary

Problems solved by technology

For example, some data owner devices with poor hardware performance may not be able to train the global business model within the specified time, or complete the global business model training within the specified time, but it needs to consume a lot of costs, such as a lot of communication costs and A lot of power consumption, etc.

Images