RT info:eu-repo/semantics/conferenceObject
T1 VQQL: a model to generalize in reinforcement learning
A1 Fernández, Fernando
A1 Borrajo, Daniel
A1 Matellán Olivera, Vicente
A2 Arquitectura y Tecnologia de Computadores
K1 Informática
K1 Conocimiento
K1 Robótica
K1 VQQL
K1 Aprendizaje
AB Reinforcement learning har proven to be very successful for finding optimal policies on uncertian and/or dynamic domains. One of the problems on using such techniques appears with large state and action spaces. This problem appears very frequently given that most information in the type of tasks to which these techniques have been applied is continuous. In the paper, we describe a new mechanism for solving the states generalization problem in reinforcement learning algorithms, the VQQL technique
YR 2012
FD 2012-10-18
LK http://hdl.handle.net/10612/1921
UL http://hdl.handle.net/10612/1921
NO European Conference on Planning, Septiembre, 1999, Durham, Reino Unido
DS BULERIA. Repositorio Institucional de la Universidad de León
RD 03-may-2024