Risk-Constrained Reinforcement Learning with Percentile Risk Criteria

zbMath1471.90160arXiv1512.01629MaRDI QIDQ4558492

Marco Pavone, Lucas Janson, Mohammad Ghavamzadeh, Yinlam Chow

Publication date: 22 November 2018

Full work available at URL: https://arxiv.org/abs/1512.01629

zbMATH Keywords

Markov decision processreinforcement learningactor-critic algorithmsconditional value-at-riskchance-constrained optimizationpolicy gradient algorithms

Mathematics Subject Classification ID

Learning and adaptive systems in artificial intelligence (68T05) Markov and semi-Markov decision processes (90C40) General considerations in statistical decision theory (62C05)

Related Items (17)

SAMBA: safe model-based \& active reinforcement learning ⋮ Risk-Sensitive Reinforcement Learning via Policy Gradient Search ⋮ On Robustness of Individualized Decision Rules ⋮ Sim-to-lab-to-real: safe reinforcement learning with shielding and generalization guarantees ⋮ Risk-averse optimization of reward-based coherent risk measures ⋮ Safety-constrained reinforcement learning with a distributional safety critic ⋮ Temporal Robustness of Stochastic Signals ⋮ Unnamed Item ⋮ Recent advances in reinforcement learning in finance ⋮ Safe multi-agent reinforcement learning for multi-robot control ⋮ Reinforcement learning with dynamic convex risk measures ⋮ Classification with costly features as a sequential decision-making problem ⋮ Global Convergence of Policy Gradient Methods to (Almost) Locally Optimal Policies ⋮ Risk-averse policy optimization via risk-neutral policy optimization ⋮ Risk verification of stochastic systems with neural network controllers ⋮ Mean-Semivariance Policy Optimization via Risk-Averse Reinforcement Learning ⋮ Peril, prudence and planning as risk, avoidance and worry

Cites Work

• Unnamed Item
• Unnamed Item
• Unnamed Item
• Unnamed Item
• Unnamed Item
• Unnamed Item
• Unnamed Item
• Unnamed Item
• Unnamed Item
• Unnamed Item
• An online actor-critic algorithm with function approximation for constrained Markov decision processes
• Stochastic recursive algorithms for optimization. Simultaneous perturbation methods
• An actor-critic algorithm with function approximation for discounted cost constrained Markov decision processes
• Dynamic mean-risk optimization in a binomial model
• Natural actor-critic algorithms
• Minimizing risk models in Markov decision processes with policies depending on target values
• Simple statistical gradient-following algorithms for connectionist reinforcement learning
• Mean, variance and probabilistic criteria in finite Markov decision processes: A review
• Markov decision processes with average-value-at-risk criteria
• Risk neutral and risk averse stochastic dual dynamic programming method
• Time consistent dynamic risk measures
• An actor-critic algorithm for constrained Markov decision processes
• Coherent Measures of Risk
• Risk-Constrained Markov Decision Processes
• Computing VaR and CVaR using stochastic approximation and adaptive unconstrained importance sampling
• Variance-Penalized Markov Decision Processes
• Multivariate stochastic approximation using a simultaneous perturbation gradient approximation
• OnActor-Critic Algorithms
• A Perturbation Theory for Ergodic Markov Chains and Application to Numerical Approximations
• Perturbation analysis for denumerable Markov chains with application to queueing models
• The variance of discounted Markov decision processes
• Percentile performance criteria for limiting average Markov decision processes
• Stochastic Target Hitting Time and the Problem of Early Retirement
• Stochastic Approximations and Differential Inclusions, Part II: Applications
• Envelope Theorems for Arbitrary Choice Sets
• Risk-Sensitive Markov Decision Processes
• Q-Learning for Risk-Sensitive Control
• A sensitivity formula for risk-sensitive cost and the actor-critic algorithm

This page was built for publication: Risk-Constrained Reinforcement Learning with Percentile Risk Criteria