conftrace_

2010 NeurIPS NeurIPS 2010

Policy gradients in linearly-solvable MDPs

Abstract

We present policy gradient results within the framework of linearly-solvable MDPs. For the first time, compatible function approximators and natural policy gradients are obtained by estimating the cost-to-go function, rather than the (much larger) state-action advantage function as is necessary in traditional MDPs. We also develop the first compatible function approximators and natural policy gradients for continuous-time stochastic systems.

🌉 Interdisciplinary Bridge - Artificial Intelligence and Mathematics & Optimization and Reinforcement Learning

📈 Trend Setter - Markov Decision Process

🐣 Hot Topic Early Bird - markov decision process

Authors

Emanuel Todorov

Topics

Machine Learning > Optimization & Theory > Optimization Reinforcement Learning > Methods > Deep RL Reinforcement Learning > Methods > Policy Learning Machine Learning > Learning Types > Reinforcement Learning Mathematics & Optimization > Optimization > Optimal Control Artificial Intelligence > Core AI > Reinforcement Learning Machine Learning > Learning Types > Markov Decision Process

Keywords

reinforcement learning policy gradient function approximation natural policy gradient markov decision process cost-to-go function linearly-solvable mdps continuous-time stochastic systems function approximator compatible function approximator linearly-solvable markov decision process

More from NeurIPS 2010

All NeurIPS 2010 papers →

Infinite Relational Modeling of Functional Connectivity in Resting State fMRI

An Alternative to Low-level-Sychrony-Based Methods for Speech Detection

A rational decision making framework for inhibitory control

Sufficient Conditions for Generating Group Level Sparsity in a Robust Minimax Framework

Learning concept graphs from text with stick-breaking priors

Fast Large-scale Mixture Modeling with Component-specific Data Partitions