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

Executing actions in a correlated manner is a common strategy for humancoordination that often leads to better cooperation, which is also potentiallybeneficial for cooperative multi-agent reinforcement learning (MARL). However,the recent success of MARL relies heavily on the convenient paradigm of purelydecentralized execution, where there is no action correlation among agents forscalability considerations. In this work, we introduce a Bayesian network toinaugurate correlations between agents' action selections in their jointpolicy. Theoretically, we establish a theoretical justification for why actiondependencies are beneficial by deriving the multi-agent policy gradient formulaunder such a Bayesian network joint policy and proving its global convergenceto Nash equilibria under tabular softmax policy parameterization in cooperativeMarkov games. Further, by equipping existing MARL algorithms with a recentmethod of differentiable directed acyclic graphs (DAGs), we develop practicalalgorithms to learn the context-aware Bayesian network policies in scenarioswith partial observability and various difficulty. We also dynamically decreasethe sparsity of the learned DAG throughout the training process, which leads toweakly or even purely independent policies for decentralized execution.Empirical results on a range of MARL benchmarks show the benefits of ourapproach.

Context-Aware Bayesian Network Actor-Critic Methods for Cooperative Multi-Agent Reinforcement Learning