AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

Continuous Action Reinforcement Learning From a Mixture of Interpretable Experts

Riad Akrour, Davide Tateo, Jan Peters

TPAMI 2021 · Reinforcement learning › Visual and representation RL · approach: Probabilistic and autonomous RL · 21 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
increase number of clusters KMetric · 7 continuous action tasksreturn: 5 -> 20 clusters approaches neural network policyImproved
use Metric mixture policy instead of linear policyTRPO linear policy · 7 continuous action tasksreturn: as little as 5 clustersImproved
use Metric instead of TRPO/PPO neural network, high-dimensional pybullet tasksTRPO/PPO · pybullet locomotion tasksreturn: matches TRPO and PPO with fewer parametersNo clear effect
use small K for interpretabilityMetric · 7 continuous action tasksreturn: outperformed by both deep RL baselines on all tasksHurt
increase K to 40 on AntBulletEnv-v0 versus baselinesMetric · AntBulletEnv-v0return: 378, 748 and 1488 parameters vs 6544Improved
select prototypical states from trajectory data instead of gradient-learned centers (ablation)Metric · pybullet locomotion tasksgait learning: none learned a gait vs walking gaits with 10 clustersImproved
add entropy constraint to differentiable prototype PPO/TRPO (ablation)Metric · pybullet locomotion tasksreturn: no gait learned -> competent policy after prototypes foundImproved
use trajectory-selected prototypes instead of differentiable ones (ablation)Metric · pybullet locomotion tasksinterpretabilityImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy