Our paper “Masked Autoencoders are Efficient Continual Federated Learners” got accepted for publication at the Conference on Lifelong Learning Agents (CoLLAs) 2024!
In the paper we investigate the challenging task of unsupervised continual learning in federated (distributed) scenarios. Our key motivation is that even if clients across a network don’t share the exact same goal, they will likely profit from the experience that other’s have made and should be able to listen to it and selectively choose beneficial parts. To this end, we show that masked autoencoders are particularly amenable to the setup. We then show the masking strategy can be integrated with task attention mechanisms and how this helps in selective knowledge transfer between clients.
Read the abstract below and the full paper for more detailed information:
“Machine learning is typically framed from a perspective of i.i.d., and more importantly, isolated data. In parts, federated learning lifts this assumption, as it sets out to solve the real-world challenge of collaboratively learning a shared model from data distributed across clients. However, motivated primarily by privacy and computational constraints, the fact that data may change, distributions drift, or even tasks advance individually on clients, is seldom taken into account. The field of continual learning addresses this separate challenge and first steps have recently been taken to leverage synergies in distributed supervised settings, in which several clients learn to solve changing classification tasks over time without forgetting previously seen ones. Motivated by these prior works, we posit that such federated continual learning should be grounded in unsupervised learning of representations that are shared across clients; in the loose spirit of how humans can indirectly leverage others’ experience without exposure to a specific task. For this purpose, we demonstrate that masked autoencoders for distribution estimation are particularly amenable to this setup. Specifically, their masking strategy can be seamlessly integrated with task attention mechanisms to enable selective knowledge transfer between clients. We empirically corroborate the latter statement through several continual federated scenarios on both image and binary datasets.”
Abstract of our paper “Masked Autoencoders are Efficient Continual Federated Learners”