Reward-weighted regression with sample reuse for direct policy search in reinforcement learning

DOI10.1162/NECO_A_00199MaRDI QIDQ2887009zbMATH OpenOpenAlexWikidataFDO

Authors Hirotaka Hachiya, Jan Peters, Masashi Sugiyama

Publication date 15 May 2012

Published in Neural Computation (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1162/neco_a_00199

Mathematics Subject Classification ID

Learning and adaptive systems in artificial intelligence (68T05)

Recommendations

• Efficient sample reuse in policy gradients with parameter-based exploration
• Reinforcement learning in sparse-reward environments with hindsight policy gradients
• Recurrent policy gradients
• Breaking the sample complexity barrier to regret-optimal model-free reinforcement learning
• scientific article; zbMATH DE number 1753141
• Expected policy gradients for reinforcement learning

Cites work

• Adaptive importance sampling for value function approximation in off-policy reinforcement learning
• Efficient exploration through active learning for value function approximation in reinforcement learning
• Improving predictive inference under covariate shift by weighting the log-likelihood function
• Input-dependent estimation of generalization error under covariate shift
• Real-time reinforcement learning by sequential actor-critics and experience replay
• Trading Variance Reduction with Unbiasedness: The Regularized Subspace Information Criterion for Robust Model Selection in Kernel Regression
• Using Expectation-Maximization for Reinforcement Learning

Cited in

(10)

• Learning under nonstationarity: covariate shift and class-balance change
• Efficient exploration through active learning for value function approximation in reinforcement learning
• Autonomous reinforcement learning with experience replay
• Importance sampling techniques for policy optimization
• Policy search for motor primitives in robotics
• Accelerating reinforcement learning with a directional-Gaussian-smoothing evolution strategy
• Efficient sample reuse in policy gradients with parameter-based exploration
• Bayesian policy reuse
• Reinforcement learning in sparse-reward environments with hindsight policy gradients
• 10.1162/jmlr.2003.3.4-5.921

Describes a project that uses

Uses Software

• Open Dynamics Engine

This page was built for publication: Reward-weighted regression with sample reuse for direct policy search in reinforcement learning

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2887009)