This article is more than 1 year old

Alibaba opens up code for federated learning platform

Claims FederatedScope is easy to use and keeps training data private

Chinese e-commerce giant Alibaba has open-sourced a federated learning platform it claims protects privacy by enabling the development of machine learning algorithms without having to share training data.

FederatedScope was developed by Alibaba's DAMO Academy, the global science and technology research outfit it founded in 2017, and the source code for this is now published under the Apache 2.0 license on GitHub.

The platform is described as a comprehensive federated learning platform that provides flexible customization for a variety of machine learning tasks in both academia and industry.

It is also claimed to be easy to get to grips with, allowing users to integrate their own components, including datasets and models for specific applications.

Federated learning, as the name suggests, is a machine learning technique that trains a model across a number of distributed nodes or hosts. Each node uses local training data, and if the model parameters are shared between nodes instead of the raw data, this means that the data itself can be kept private.

According to Alibaba, gathering training data to build and evolve machine learning models is increasingly coming under scrutiny because of the potential privacy concerns, and federated learning can help address some of these concerns.

"By sharing our self-developed federated learning technologies with the open-source community, we hope to promote the research and industrial deployment of privacy-preserving computation in different sectors, such as healthcare and smart mobility that usually involves sensitive user data and requires strict privacy protection practices," DAMO Academy research scientist Bolin Ding said in a statement.

FederatedScope features an event-driven architecture and includes various tools such as a collection of benchmark datasets, well-known model architectures, sample federated learning algorithms, and automated tuning functions, Alibaba said.

These capabilities enable developers to build and customize task-specific federated learning applications targeting areas such as computer vision, natural language processing, speech recognition, graph learning, and recommendation.

FederatedScope also offers privacy protection through the use of differential privacy and multi-party computation to meet different requirements of privacy protection, Alibaba said.

Alibaba is not the only organization making federated learning tools available. Last month, HPE introduced Swarm Learning, its own decentralized machine learning framework for edge applications or distributed sites.

HPE Swarm Learning is provided as part of a Swarm Learning Library that is containerized and can run on Docker, inside virtual machines or on bare metal, and uses blockchain technology to ensure that model parameters can be exchanged securely. ®

More about


Send us news

Other stories you might like