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/

Zendy Plus

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

Zendy Tools

Zendy Open

Many recent breakthroughs in multi-agent reinforcement learning (MARL)require the use of deep neural networks, which are challenging for humanexperts to interpret and understand. On the other hand, existing work oninterpretable reinforcement learning (RL) has shown promise in extracting moreinterpretable decision tree-based policies from neural networks, but only inthe single-agent setting. To fill this gap, we propose the first set ofalgorithms that extract interpretable decision-tree policies from neuralnetworks trained with MARL. The first algorithm, IVIPER, extends VIPER, arecent method for single-agent interpretable RL, to the multi-agent setting. Wedemonstrate that IVIPER learns high-quality decision-tree policies for eachagent. To better capture coordination between agents, we propose a novelcentralized decision-tree training algorithm, MAVIPER. MAVIPER jointly growsthe trees of each agent by predicting the behavior of the other agents usingtheir anticipated trees, and uses resampling to focus on states that arecritical for its interactions with other agents. We show that both algorithmsgenerally outperform the baselines and that MAVIPER-trained agents achievebetter-coordinated performance than IVIPER-trained agents on three differentmulti-agent particle-world environments.

MAVIPER: Learning Decision Tree Policies for Interpretable Multi-Agent Reinforcement Learning