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 Plus annual

Presents the access of premium content as premium feature

Premium Content

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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Value function factorization methods have become a dominant approach forcooperative multiagent reinforcement learning under a centralized training anddecentralized execution paradigm. By factorizing the optimal joint action-valuefunction using a monotonic mixing function of agents' utilities, thesealgorithms ensure the consistency between joint and local action selections fordecentralized decision-making. Nevertheless, the use of monotonic mixingfunctions also induces representational limitations. Finding the optimalprojection of an unrestricted mixing function onto monotonic function classesis still an open problem. To this end, we propose ReMIX, formulating thisoptimal projection problem for value function factorization as a regretminimization over the projection weights of different state-action values. Suchan optimization problem can be relaxed and solved using the Lagrangianmultiplier method to obtain the close-form optimal projection weights. Byminimizing the resulting policy regret, we can narrow the gap between theoptimal and the restricted monotonic mixing functions, thus obtaining animproved monotonic value function factorization. Our experimental results onPredator-Prey and StarCraft Multiagent Challenge environments demonstrate theeffectiveness of our method, indicating the better capabilities of handlingenvironments with non-monotonic value functions.

ReMIX: Regret Minimization for Monotonic Value Function Factorization in Multiagent Reinforcement Learning