FedLess - Secure and Scalable Serverless Federated Learning
Explore the innovative FedLess framework in this 31-minute conference talk by Mohak Chadha from the Technical University of Munich. Delve into the world of federated learning (FL) and discover how serverless computing addresses its challenges. Learn about FedLess, an open-source cloud-agnostic framework that enables scalable FL for training arbitrary DNN models using serverless platforms like AWS Lambda, Google Cloud Functions, and Azure functions. Examine built-in strategies to tackle common issues in scalable FL, such as stragglers and resource heterogeneity among clients. Witness a live demonstration of setting up a client federation for training using FedLess. Gain valuable insights into the intersection of federated learning and serverless computing, making this talk essential for those interested in advancing their knowledge in these cutting-edge fields.