Distributed Continual Learning

TL;DR

This paper introduces a mathematical framework for distributed continual learning, analyzing different information sharing modes, and demonstrates that modular parameter sharing offers optimal performance with minimal communication.

Contribution

It presents a comprehensive framework for distributed continual learning, develops algorithms for various sharing modes, and empirically evaluates their efficiency and effectiveness.

Findings

Sharing parameters is more efficient than sharing data for complex tasks.

Modular parameter sharing achieves the best performance with low communication costs.

Combining sharing modes can further improve learning performance.

Abstract

This work studies the intersection of continual and federated learning, in which independent agents face unique tasks in their environments and incrementally develop and share knowledge. We introduce a mathematical framework capturing the essential aspects of distributed continual learning, including agent model and statistical heterogeneity, continual distribution shift, network topology, and communication constraints. Operating on the thesis that distributed continual learning enhances individual agent performance over single-agent learning, we identify three modes of information exchange: data instances, full model parameters, and modular (partial) model parameters. We develop algorithms for each sharing mode and conduct extensive empirical investigations across various datasets, topology structures, and communication limits. Our findings reveal three key insights: sharing parameters is more efficient than sharing data as tasks become more complex; modular parameter sharing yields the best performance while minimizing communication costs; and combining sharing modes can cumulatively improve performance.