Transverse federated learning method, device and equipment and computer storage medium

Abstract

The invention relates to the technical field of Fintech, and discloses a transverse federated learning method. The method comprises the following steps: a community coordinator acquiring global model parameters sent by a central coordinator and sending the global model parameters to each participant; obtaining model parameter update which is sent by each participant and is obtained by performing model training based on the global model parameters, fusing the model parameter update to obtain community model parameter update, and determining whether the community model parameter update needs to be sent to the central coordinator or not; and if yes, sending the community model parameter update to the central coordinator, obtaining global model parameter update returned by the central coordinator, and sending the global model parameter update to each participant, thereby enabling each participant to carry out model training based on the global model parameter update. The invention further discloses a transverse federated learning device and equipment and a computer storage medium. According to the invention, the learning efficiency of transverse federal learning is improved.

Description

technical field [0001] The present invention relates to the technical field of artificial intelligence financial technology (Fintech), in particular to a horizontal federated learning method, device, equipment and computer storage medium. Background technique [0002] With the development of computer technology, more and more technologies (big data, distributed, blockchain, artificial intelligence, etc.) The industry's security and real-time requirements also put forward higher requirements for technology. For example, federated learning is a method of machine learning by combining different participants, and horizontal federated learning is when the data characteristics of participants (such as mobile terminals or Internet of Things devices) overlap more and users overlap less Next, take out the part of the data with the same characteristics of the participants' user data but not the same users for joint machine learning. But for IoT devices, in addition to the very limit...

AI Technical Summary

Problems solved by technology

But for IoT devices, in addition to the very limited communication bandwidth, power and computing power are also very limited

If it is necessary to perform encryption operations on the model parameter updates during the horizontal federated learning training process, the encryption operations involved (especially the corresponding decryption operations) may exceed the power and computing power of the IoT device, thereby affecting the available encryption. Complexity and Secrecy Effects

For mobile terminals (for example, smart phones), encryption and decryption operations may also consume too much power of the mobile terminal, thereby affecting user experience

Images