Bayesian inverse reinforcement learning
作者： Eyal Amir , Deepak Ramachandran
DOI:
关键词: Probability distribution 、 Reward learning 、 Generalization error 、 Active learning (machine learning) 、 Machine learning 、 Temporal difference learning 、 Markov decision process 、 Unsupervised learning 、 Stability (learning theory) 、 Learning classifier system 、 Semi-supervised learning 、 Q-learning 、 Apprenticeship learning 、 Reinforcement learning 、 Preference elicitation 、 Instance-based learning 、 Preference learning 、 Artificial intelligence 、 Heuristic 、 Computer science
摘要: Inverse Reinforcement Learning (IRL) is the problem of learning reward function underlying a Markov Decision Process given dynamics system and behaviour an expert. IRL motivated by situations where knowledge rewards goal itself (as in preference elicitation) task apprenticeship (learning policies from expert). In this paper we show how to combine prior evidence expert's actions derive probability distribution over space functions. We present efficient algorithms that find solutions for tasks generalize well these distributions. Experimental results strong improvement our methods previous heuristic-based approaches.
ijcai.org
acm.org 参考文章(15)
Albert Tarantola, Inverse Problem Theory and Methods for Model Parameter Estimation ,(2004)
Damien Ernst, Arthur Louette, Introduction to Reinforcement Learning MIT Press. ,(1998)
Craig Boutilier, Bob Price, A Bayesian approach to imitation in reinforcement learning international joint conference on artificial intelligence. pp. 712- 717 ,(2003)
Stefan Schaal, Christopher G. Atkeson, Robot Learning From Demonstration international conference on machine learning. pp. 12- 20 ,(1997)
Santosh Vempala, Geometric Random Walks: a Survey Combinatorial and Computational Geometry, 2007, ISBN 0-521-84862-8, págs. 577-616. pp. 577- 616 ,(2007)
James O Berger, Statistical Decision Theory and Bayesian Analysis ,(1993)
Pieter Abbeel, Andrew Y. Ng, Apprenticeship learning via inverse reinforcement learning Twenty-first international conference on Machine learning - ICML '04. pp. 1- 8 ,(2004) , 10.1145/1015330.1015430
Stuart Russell, Learning agents for uncertain environments (extended abstract) conference on learning theory. pp. 101- 103 ,(1998) , 10.1145/279943.279964
Andrew Y Ng, Stuart Russell, None, Algorithms for Inverse Reinforcement Learning international conference on machine learning. ,vol. 67, pp. 663- 670 ,(2000) , 10.2460/AJVR.67.2.323
David Applegate, Ravi Kannan, Sampling and integration of near log-concave functions symposium on the theory of computing. pp. 156- 163 ,(1991) , 10.1145/103418.103439