Classical Policy Gradient: Preserving Bellman's Principle of Optimality

06/06/2019

∙

We propose a new objective function for finite-horizon episodic Markov decision processes that better captures Bellman's principle of optimality, and provide an expression for the gradient of the objective.

READ FULL TEXT

Classical Policy Gradient: Preserving Bellman's Principle of Optimality

Policy-based Primal-Dual Methods for Convex Constrained Markov Decision Processes

Global optimality of softmax policy gradient with single hidden layer neural networks in the mean-field regime

Trust Region Policy Optimization of POMDPs

Policy Gradient Algorithms for Robust MDPs with Non-Rectangular Uncertainty Sets

On the Convergence and Optimality of Policy Gradient for Markov Coherent Risk

Edge Rewiring Goes Neural: Boosting Network Resilience via Policy Gradient

A Policy Gradient Method for Confounded POMDPs

Classical Policy Gradient: Preserving Bellman's Principle of Optimality

Related Research

Policy-based Primal-Dual Methods for Convex Constrained Markov Decision Processes

Global optimality of softmax policy gradient with single hidden layer neural networks in the mean-field regime

Trust Region Policy Optimization of POMDPs

Policy Gradient Algorithms for Robust MDPs with Non-Rectangular Uncertainty Sets

On the Convergence and Optimality of Policy Gradient for Markov Coherent Risk

Edge Rewiring Goes Neural: Boosting Network Resilience via Policy Gradient

A Policy Gradient Method for Confounded POMDPs