A stochastic trust-region framework for policy optimization

DOI10.4208/JCM.2104-M2021-0007MaRDI QIDQ5096136OpenAlexFDO

Authors Mingming Zhao, Yongfeng Li, Zaiwen Wen

Publication date 15 August 2022

Published in Journal of Computational Mathematics (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1911.11640

zbMATH Keywords

global convergencedeep reinforcement learningpolicy optimizationentropy controlstochastic trust region method

Mathematics Subject Classification ID

Nonconvex programming, global optimization (90C26) Stochastic programming (90C15) Markov and semi-Markov decision processes (90C40) Optimal stochastic control (93E20)

Recommendations

• Expected policy gradients for reinforcement learning
• Compatible natural gradient policy search
• Approximate Newton Policy Gradient Algorithms
• Stochastic trust-region methods with trust-region radius depending on probabilistic models
• A generalized path integral control approach to reinforcement learning

Cites work

• scientific article; zbMATH DE number 5060482 (Why is no real title available?)
• Approximate Newton methods for policy search in Markov decision processes
• Convergence of trust-region methods based on probabilistic models
• Global convergence of policy gradient methods to (almost) locally optimal policies
• Optimization theory and methods. Nonlinear programming
• Stochastic optimization using a trust-region method and random models

Cited in

(2)

• scientific article; zbMATH DE number 6982305 (Why is no real title available?)
• A Stochastic Composite Augmented Lagrangian Method for Reinforcement Learning

Describes a project that uses

Uses Software

• OpenAI Gym

This page was built for publication: A stochastic trust-region framework for policy optimization

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5096136)