当前位置:
X-MOL 学术
›
arXiv.cs.MA
›
论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Scaling Multi-Agent Reinforcement Learning with Selective Parameter Sharing
arXiv - CS - Multiagent Systems Pub Date : 2021-02-15 , DOI: arxiv-2102.07475 Filippos Christianos, Georgios Papoudakis, Arrasy Rahman, Stefano V. Albrecht
arXiv - CS - Multiagent Systems Pub Date : 2021-02-15 , DOI: arxiv-2102.07475 Filippos Christianos, Georgios Papoudakis, Arrasy Rahman, Stefano V. Albrecht
Sharing parameters in multi-agent deep reinforcement learning has played an
essential role in allowing algorithms to scale to a large number of agents.
Parameter sharing between agents significantly decreases the number of
trainable parameters, shortening training times to tractable levels, and has
been linked to more efficient learning. However, having all agents share the
same parameters can also have a detrimental effect on learning. We demonstrate
the impact of parameter sharing methods on training speed and converged
returns, establishing that when applied indiscriminately, their effectiveness
is highly dependent on the environment. Therefore, we propose a novel method to
automatically identify agents which may benefit from sharing parameters by
partitioning them based on their abilities and goals. Our approach combines the
increased sample efficiency of parameter sharing with the representational
capacity of multiple independent networks to reduce training time and increase
final returns.
更新日期:2021-02-16