Distributed federated learning cooperative computing method and system

Abstract

The invention discloses a distributed federated learning cooperative computing method and system. The distributed federated learning cooperative computing method specifically comprises the following steps: performing deep reinforcement learning model training; in response to deployment of the trained deep reinforcement learning model to each edge server, carrying out federated learning; and ending the federal learning. According to the invention, for a distributed federated learning framework, the dependence of traditional federated learning on a central server is broken, and the privacy protection and security of the federated learning process are effectively ensured.

Description

technical field [0001] The present application relates to the communication field, and in particular to a distributed federated learning collaborative computing method and system. Background technique [0002] Metal material workpieces are an important part of some products in the machining process. The quality of metal material workpieces directly affects the market competitiveness of enterprise products. Therefore, it is very important to detect the surface defects of metal material workpieces in the machining process. . For the defect detection of metal surfaces, deep learning technology can be used to collect workpiece images from the production line, and then extract defect information from the images. By learning the surface defect characteristics of metal workpieces, a network detection and defect recognition model for metal workpiece defects can be established. Commonly used detection models include Fast R-CNN, Faster R-CNN, Mask R-CNN, etc. However, in the industr...

AI Technical Summary

Problems solved by technology

[0006] Based on this, this application provides a distributed federated learning collaborative computing method for smart factories, which ensures the safety of the federated learning process, and uses deep reinforcement learning (deep reinforcement learning, DRL) technology to solve the problem between edge servers and participants. The association and bandwidth resource allocation problem and the computing resource allocation problem of the participants

Images