English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Traditional centralized multi-agent reinforcement learning (MARL) algorithmsare sometimes unpractical in complicated applications, due to non-interactivitybetween agents, curse of dimensionality and computation complexity. Hence,several decentralized MARL algorithms are motivated. However, existingdecentralized methods only handle the fully cooperative setting where massiveinformation needs to be transmitted in training. The block coordinate gradientdescent scheme they used for successive independent actor and critic steps cansimplify the calculation, but it causes serious bias. In this paper, we proposea flexible fully decentralized actor-critic MARL framework, which can combinemost of actor-critic methods, and handle large-scale general cooperativemulti-agent setting. A primal-dual hybrid gradient descent type algorithmframework is designed to learn individual agents separately fordecentralization. From the perspective of each agent, policy improvement andvalue evaluation are jointly optimized, which can stabilize multi-agent policylearning. Furthermore, our framework can achieve scalability and stability forlarge-scale environment and reduce information transmission, by the parametersharing mechanism and a novel modeling-other-agents methods based ontheory-of-mind and online supervised learning. Sufficient experiments incooperative Multi-agent Particle Environment and StarCraft II show that ourdecentralized MARL instantiation algorithms perform competitively againstconventional centralized and decentralized methods.

F2A2: Flexible Fully-decentralized Approximate Actor-critic for Cooperative Multi-agent Reinforcement Learning