MARCO: A Cooperative Knowledge Transfer Framework for Personalized Cross-domain Recommendations

2025-10-06

AI Summary

To address negative transfer in cross-domain recommendation—caused by cold-start and data sparsity issues stemming from imbalanced source-domain contributions and distributional discrepancies—this paper proposes a Multi-Agent Reinforcement Learning-based Knowledge Transfer framework (MAKT). MAKT models each source domain as an independent agent and employs a cooperative credit assignment mechanism to dynamically optimize domain-specific contribution weights. It further introduces an entropy-driven action diversity penalty to enhance policy expressiveness and training stability, and integrates cross-domain embedding alignment to ensure effective knowledge transfer. Extensive experiments on four benchmark datasets demonstrate that MAKT significantly outperforms existing single-agent and conventional cross-domain recommendation methods, achieving superior generalization and robustness under varying data conditions. The implementation code is publicly available.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Multiagent Systems: Adversarial Agents
• Application Domains: Humanities & Computational Social Science

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Recommender systems frequently encounter data sparsity issues, particularly when addressing cold-start scenarios involving new users or items. Multi-source cross-domain recommendation (CDR) addresses these challenges by transferring valuable knowledge from multiple source domains to enhance recommendations in a target domain. However, existing reinforcement learning (RL)-based CDR methods typically rely on a single-agent framework, leading to negative transfer issues caused by inconsistent domain contributions and inherent distributional discrepancies among source domains. To overcome these limitations, MARCO, a Multi-Agent Reinforcement Learning-based Cross-Domain recommendation framework, is proposed. It leverages cooperative multi-agent reinforcement learning, where each agent is dedicated to estimating the contribution from an individual source domain, effectively managing credit assignment and mitigating negative transfer. In addition, an entropy-based action diversity penalty is introduced to enhance policy expressiveness and stabilize training by encouraging diverse agents' joint actions. Extensive experiments across four benchmark datasets demonstrate MARCO's superior performance over state-of-the-art methods, highlighting its robustness and strong generalization capabilities. The code is at https://github.com/xiewilliams/MARCO.

Problem

• Addressing data sparsity in cold-start recommender systems
• Mitigating negative transfer in cross-domain recommendation
• Managing inconsistent domain contributions via multi-agent learning

Innovation

• Multi-agent reinforcement learning for cross-domain recommendation
• Entropy-based action diversity penalty stabilizes training
• Individual agents estimate contributions from source domains