conftrace_

2010 NeurIPS NeurIPS 2010

Nonparametric Bayesian Policy Priors for Reinforcement Learning

Abstract

We consider reinforcement learning in partially observable domains where the agent can query an expert for demonstrations. Our nonparametric Bayesian approach combines model knowledge, inferred from expert information and independent exploration, with policy knowledge inferred from expert trajectories. We introduce priors that bias the agent towards models with both simple representations and simple policies, resulting in improved policy and model learning.

🌉 Interdisciplinary Bridge - Artificial Intelligence and Reinforcement Learning

📈 Trend Setter - Imitation Learning

🐣 Hot Topic Early Bird - partially observable markov decision process

Authors

Finale Doshi-velez, David Wingate, Nicholas Roy, Joshua B. Tenenbaum

Topics

Artificial Intelligence > Bayesian & Probabilistic > Bayesian Learning Machine Learning > Learning Types > Self-Supervised Learning Reinforcement Learning > Methods > Deep RL Reinforcement Learning > Methods > Policy Learning Machine Learning > Bayesian & Probabilistic > Bayesian Learning Machine Learning > Learning Types > Imitation Learning Machine Learning > Bayesian & Probabilistic > Nonparametric Bayesian

Keywords

reinforcement learning nonparametric bayesian policy learning partially observable domains partially observable markov decision process partially observable expert demonstration model learning policy prior

More from NeurIPS 2010

All NeurIPS 2010 papers →

Parametric Bandits: The Generalized Linear Case

Basis Construction from Power Series Expansions of Value Functions

A Novel Kernel for Learning a Neuron Model from Spike Train Data

On the Theory of Learnining with Privileged Information

Accounting for network effects in neuronal responses using L1 regularized point process models

Probabilistic latent variable models for distinguishing between cause and effect