Título: The Essential Dynamics Algorithm: Essential Results
Autores: Martin, Martin C.
Fecha: 2004-10-08
2004-10-08
2003-05-01
Publicador: MIT
Fuente:
Tipo:
Tema: AI
Reinforcement learning
bicycle
policy search
markov decision processes
Descripción: This paper presents a novel algorithm for learning in a class of stochastic Markov decision processes (MDPs) with continuous state and action spaces that trades speed for accuracy. A transform of the stochastic MDP into a deterministic one is presented which captures the essence of the original dynamics, in a sense made precise. In this transformed MDP, the calculation of values is greatly simplified. The online algorithm estimates the model of the transformed MDP and simultaneously does policy search against it. Bounds on the error of this approximation are proven, and experimental results in a bicycle riding domain are presented. The algorithm learns near optimal policies in orders of magnitude fewer interactions with the stochastic MDP, using less domain knowledge. All code used in the experiments is available on the project's web site.
Idioma: Inglés

Artículos similares:

Description Of Procedures In Automotive Engine Plants por Artzner, Denis,Whitney, Dr. Daniel
Reading Courtesy Amounts on Handwritten Paper Checks por Palacios, Rafael,Wang, Patrick S.P.,Gupta, Amar
On Trees and Logs por Pavlova, Anna,Cass, David
Saturn, The GM/UAW Partnership por Rubinstein, Saul,Kochan, Thomas
Academic Earmarks and the Returns to Lobbying por De Figueiredo, John M.,Silverman, Brian S.
10