Title :
Automatic tuning of judgement parameter in continuous state exploitation-oriented learning
Author :
Miyazaki, Kazuteru
Author_Institution :
Dept. of Assessment & Res. for Degree Awarding, Univ. Evaluation, Tokyo, Japan
Abstract :
The rational policy making algorithm (PPM) and the penalty avoiding rational policy making algorithm (PARP) under continuous state spaces has important parameter that decides the same of basic functions. It is necessary to set an appropriate value through a preliminary experiment. In this paper, we propose an automatic tuning mechanism of the judgement parameter. We show the effectiveness of our proposal using a pole-cart problem.
Keywords :
decision making; learning (artificial intelligence); automatic tuning mechanism; continuous state exploitation oriented learning; judgement parameter; penalty avoiding rational policy making algorithm; pole-cart problem; Algorithm design and analysis; Learning; Machine learning; Markov processes; Proposals; Tuning; Continuous State Spaces; Exploitation-oriented Learning XoL; PARP; RPM; Reinforcement Learning;
Conference_Titel :
SICE Annual Conference 2010, Proceedings of
Conference_Location :
Taipei
Print_ISBN :
978-1-4244-7642-8
