Safe-EF: Error Feedback for Nonsmooth Constrained Optimization

TL;DR

Safe-EF introduces a new error feedback algorithm for non-smooth constrained optimization in federated learning, improving communication efficiency and safety in practical distributed training scenarios.

Contribution

It establishes lower bounds for contractive compression in non-smooth convex optimization and proposes Safe-EF, a method that matches these bounds while enforcing safety constraints.

Findings

Safe-EF effectively reduces communication in federated learning.

It ensures safety constraints in distributed optimization.

Experimental results validate improved performance in reinforcement learning tasks.

Abstract

Federated learning faces severe communication bottlenecks due to the high dimensionality of model updates. Communication compression with contractive compressors (e.g., Top-K) is often preferable in practice but can degrade performance without proper handling. Error feedback (EF) mitigates such issues but has been largely restricted for smooth, unconstrained problems, limiting its real-world applicability where non-smooth objectives and safety constraints are critical. We advance our understanding of EF in the canonical non-smooth convex setting by establishing new lower complexity bounds for first-order algorithms with contractive compression. Next, we propose Safe-EF, a novel algorithm that matches our lower bound (up to a constant) while enforcing safety constraints essential for practical applications. Extending our approach to the stochastic setting, we bridge the gap between theory and practical implementation. Extensive experiments in a reinforcement learning setup, simulating distributed humanoid robot training, validate the effectiveness of Safe-EF in ensuring safety and reducing communication complexity.