Federal learning method, system and device based on hierarchical and fragmented block chain

Abstract

The invention relates to the technical field of block chains and federated learning, and relates to a federal learning method, system and device based on hierarchical and fragmented block chain, and the method comprises the steps that a main chain is constructed and a plurality of sub-chains are triggered through employing an intelligent contract; each sub-chain obtains task information and a basic model from the main chain, packages the task information and the basic model into a sub-chain transaction and submits the sub-chain transaction to the sub-chain; the intelligent terminal selected by each sub-chain main node pulls the task information and the basic model from the corresponding sub-chain for multi-round training; the sub-chain main node aggregates the sub-chain transactions uploaded by the intelligent terminals to obtain a fragmentation model of the round, and packages the fragmentation model into sub-chain transactions and submits the sub-chain transactions to the sub-chain; and each sub-chain main node packages the last round of fragmentation model in each fragmentation iteration into a main chain transaction, submits the main chain transaction to a main chain, and approves part of main chain transactions in the main chain. According to the hierarchical and fragmented block chain structure provided by the invention, the transaction processing burden caused by the fact that a single block chain network bears the whole federated learning task is reduced, and the multi-dimensional gain caused by the fusion of the block chain technology and federated learning is brought into full play.

Description

technical field [0001] The present invention relates to the technical fields of blockchain and federated learning, in particular to a federated learning method, system and device based on layered and fragmented blockchains. Background technique [0002] Federated learning is a machine learning framework. Due to its distributed characteristics, each training participant (terminal or organization) can coordinate all parties to train the same model without uploading local original data, and then meet the requirements of user privacy protection, government Complete information sharing under the requirements of laws and regulations. [0003] In order to solve the security and efficiency problems in traditional federated learning, blockchain technology is introduced into the federated learning architecture in the existing technology to realize decentralized training. However, the existing blockchain-based federated learning fails to fully consider the huge demand for computing an...

AI Technical Summary

Problems solved by technology

[0005] Embodiments of the present invention provide a federated learning method, system, and device based on layered and fragmented blockchains to solve the problem of insufficient consideration of the huge demand for computing and storage resources of some consensus algorithms in federated learning. The throughput is also difficult to support large-scale training tasks and other issues

Images