Prism: Spectral Parameter Sharing for Multi-Agent Reinforcement Learning

TL;DR

Prism introduces a spectral domain parameter sharing method in multi-agent reinforcement learning that fosters diversity among agents while maintaining scalability and resource efficiency.

Contribution

The paper presents Prism, a novel spectral parameter sharing framework using SVD to induce diversity among agents without sacrificing scalability.

Findings

Achieves competitive performance on multiple benchmarks.

Maintains resource efficiency compared to existing methods.

Fosters inter-agent diversity effectively.

Abstract

Parameter sharing is a key strategy in multi-agent reinforcement learning (MARL) for improving scalability, yet conventional fully shared architectures often collapse into homogeneous behaviors. Recent methods introduce diversity through clustering, pruning, or masking, but typically compromise resource efficiency. We propose Prism, a parameter sharing framework that induces inter-agent diversity by representing shared networks in the spectral domain via singular value decomposition (SVD). All agents share the singular vector directions while learning distinct spectral masks on singular values. This mechanism encourages inter-agent diversity and preserves scalability. Extensive experiments on both homogeneous (LBF, SMACv2) and heterogeneous (MaMuJoCo) benchmarks show that Prism achieves competitive performance with superior resource efficiency.