Federal learning training method, system and device based on block chain and storage medium

The invention provides a federal learning training method, system and device based on a block chain and a storage medium, local client nodes are added to form a block chain network, local model gradients trained by the local client nodes are aggregated on the block chain network nodes according to cooperation of multiple smart contracts, and the local model gradients of the local client nodes are aggregated on the block chain network nodes. Finally, a decentralized and efficient federal learning training method is realized, and the problems of single-point failure and incapability of verifying and auditing training data in the prior art are solved. 本发明提供一种基于区块链的联邦学习训练方法、系统、设备及存储介质，通过将本地客户端节点加入组成区块链网络，在根据多种智能合约的协同，将本地客户端节点训练的局部模型梯度在区块链网络节点上进行聚合，最终实现去中心化的、高效的联邦学习训练方法，克服了现有技术存在的单点故障问题及训练数据无法验证审核的问题。