Personalized Collaborative Learning with Affinity-Based Variance Reduction

2025-10-17

AI Summary

Addressing the challenge of simultaneously achieving collaborative learning and personalization in heterogeneous multi-agent systems, this paper proposes the Personalized Collaborative Learning (PCL) framework. PCL requires no prior knowledge of heterogeneity and adaptively balances collaboration gains against individual adaptation via bias correction, importance weighting, and affinity-aware gradient updates. A key innovation is the introduction of an affinity-based variance reduction technique, which— for the first time—enables smooth interpolation between federated linear speedup and the independent learning baseline, while revealing a novel mechanism by which acceleration remains attainable even under high heterogeneity. Theoretically, PCL achieves a sample complexity reduced to $max{n^{-1}, delta}$ times that of independent learning, ensuring both efficiency and robustness across varying degrees of heterogeneity.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Multiagent Systems: Multiagent Learning
• Search and Optimization: Learning to Search

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Multi-agent learning faces a fundamental tension: leveraging distributed collaboration without sacrificing the personalization needed for diverse agents. This tension intensifies when aiming for full personalization while adapting to unknown heterogeneity levels -- gaining collaborative speedup when agents are similar, without performance degradation when they are different. Embracing the challenge, we propose personalized collaborative learning (PCL), a novel framework for heterogeneous agents to collaboratively learn personalized solutions with seamless adaptivity. Through carefully designed bias correction and importance correction mechanisms, our method AffPCL robustly handles both environment and objective heterogeneity. We prove that AffPCL reduces sample complexity over independent learning by a factor of $max{n^{-1}, δ}$, where $n$ is the number of agents and $δin[0,1]$ measures their heterogeneity. This affinity-based acceleration automatically interpolates between the linear speedup of federated learning in homogeneous settings and the baseline of independent learning, without requiring prior knowledge of the system. Our analysis further reveals that an agent may obtain linear speedup even by collaborating with arbitrarily dissimilar agents, unveiling new insights into personalization and collaboration in the high heterogeneity regime.

Problem

• Addresses multi-agent learning tension between collaboration and personalization needs
• Proposes collaborative framework for heterogeneous agents with adaptive personalization
• Reduces sample complexity through affinity-based variance reduction mechanisms

Innovation

• Personalized collaborative learning with bias correction
• Affinity-based variance reduction for heterogeneous agents
• Automatic interpolation between federated and independent learning