天翼联邦学习平台

天翼联邦学习平台可以帮助大数据建模厂商在新的《数安法》合规要求下，建模各参与方在数据不互通、不出私域的情况下进行联合机器学习建模，各方只持有自己相关部分的模型，梯度以密文的方式汇总，避免暴露。经过多轮迭代学习最终聚合出结果模型。

The Tianyi Federated Learning Platform assists big data modeling vendors in meeting the compliance requirements of the newly enacted Data Security Law of the People's Republic of China. It enables all participating modeling parties to carry out federated machine learning modeling without data intercommunication and without allowing data to leave their private domains. Each participating party only holds the model segment corresponding to its own data, and gradients are aggregated in an encrypted manner to avoid exposing sensitive information. After multiple rounds of iterative learning, the final result model is obtained through aggregation.

搜集汇总

背景与挑战

背景概述

该平台为大数据建模厂商提供合规的联邦学习解决方案，支持参与方在数据隔离条件下通过加密梯度交互进行联合建模，经多轮迭代后生成最终模型。