Alibaba Group’s global research initiative (Alibaba DAMO Academy) has made the source code of its learning platform FederatedScope available to the open source community. […]
In so-called federated learning, the training of micro tasks is coordinated across different devices. Only the intermediate results of the training – and not the raw data of the user – are returned to the cloud server. Bolin Ding, Research Scientist at the Alibaba DAMO Academy: “By sharing our self-developed federated learning technologies with the open source community, we want to promote the research and industrial use of privacy-friendly calculations in various sectors. Such as in the healthcare sector or in the field of intelligent mobility, which usually contain sensitive user data and require strict data protection practices“.
In addition, with a newly implemented event-driven framework, FederatedScope offers various tools, including a collection of benchmark datasets, well-known model architectures, algorithms for federated learning, easy-to-use automatic tuning functions and user-friendly interfaces. These enable researchers and developers to quickly create and customize task-specific collaborative learning applications in areas such as computer vision, natural language processing, speech recognition, graph learning, and recommendations.