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

We propose FACtored Multi-Agent Centralised policy gradients (FACMAC), a newmethod for cooperative multi-agent reinforcement learning in both discrete andcontinuous action spaces. Like MADDPG, a popular multi-agent actor-criticmethod, our approach uses deep deterministic policy gradients to learnpolicies. However, FACMAC learns a centralised but factored critic, whichcombines per-agent utilities into the joint action-value function via anon-linear monotonic function, as in QMIX, a popular multi-agent Q-learningalgorithm. However, unlike QMIX, there are no inherent constraints on factoringthe critic. We thus also employ a nonmonotonic factorisation and empiricallydemonstrate that its increased representational capacity allows it to solvesome tasks that cannot be solved with monolithic, or monotonically factoredcritics. In addition, FACMAC uses a centralised policy gradient estimator thatoptimises over the entire joint action space, rather than optimising over eachagent's action space separately as in MADDPG. This allows for more coordinatedpolicy changes and fully reaps the benefits of a centralised critic. Weevaluate FACMAC on variants of the multi-agent particle environments, a novelmulti-agent MuJoCo benchmark, and a challenging set of StarCraft IImicromanagement tasks. Empirical results demonstrate FACMAC's superiorperformance over MADDPG and other baselines on all three domains.

FACMAC: Factored Multi-Agent Centralised Policy Gradients