当研究室で取り組んでいる研究に関する論文・講演などの情報（一部抜粋）です。
Here is a selection of our published papers on the research we’ve been working on in our laboratory.
Our Selected Papers and Keywords
2005
Kengo Katayama, Takahiro Koshiishi, Hiroyuki Narihisa
Reinforcement learning agents with primary knowledge designed by analytic hierarchy process Proceedings Article
In: SAC 2005: Proceedings of the 2005 ACM symposium on Applied computing, pp. 14-21, 2005.
Abstract | BibTeX | タグ: Analytic hierarchy process, Machine learning, Profit-sharing, Reinforcement learning, Sokoban problem | Links:
@inproceedings{KATAYAMA-SAC-2005,
title = {Reinforcement learning agents with primary knowledge designed by analytic hierarchy process},
author = {Kengo Katayama and Takahiro Koshiishi and Hiroyuki Narihisa},
doi = {https://doi.org/10.1145/1066677.1066685},
year = {2005},
date = {2005-03-13},
urldate = {2005-03-13},
booktitle = {SAC 2005: Proceedings of the 2005 ACM symposium on Applied computing},
pages = {14-21},
abstract = {This paper presents a novel model of reinforcement learning agents. A feature of our learning agent model is to integrate analytic hierarchy process (AHP) into a standard reinforcement learning agent model, which consists of three modules: state recognition, learning, and action selecting modules. In our model, AHP module is designed with primary knowledge that human intrinsically should have in order to attain a goal state. This aims at increasing promising actions of agent especially in the earlier stages of learning instead of completely random actions as in the standard reinforcement learning algorithms. We adopt profit-sharing as a reinforcement learning algorithm and demonstrate the potential of our approach on two learning problems of a pursuit problem and a Sokoban problem with deadlock in the grid-world domains, where results indicate that the learning time can be decreased considerably for the problems and our approach efficiently avoids the deadlock for the Sokoban problem. We also show that bad effect that can be usually observed by introducing a priori knowledge into reinforcement learning process can be restrained by a method that decreases a rate of using knowledge during learning.},
keywords = {Analytic hierarchy process, Machine learning, Profit-sharing, Reinforcement learning, Sokoban problem},
pubstate = {published},
tppubtype = {inproceedings}
}
This paper presents a novel model of reinforcement learning agents. A feature of our learning agent model is to integrate analytic hierarchy process (AHP) into a standard reinforcement learning agent model, which consists of three modules: state recognition, learning, and action selecting modules. In our model, AHP module is designed with primary knowledge that human intrinsically should have in order to attain a goal state. This aims at increasing promising actions of agent especially in the earlier stages of learning instead of completely random actions as in the standard reinforcement learning algorithms. We adopt profit-sharing as a reinforcement learning algorithm and demonstrate the potential of our approach on two learning problems of a pursuit problem and a Sokoban problem with deadlock in the grid-world domains, where results indicate that the learning time can be decreased considerably for the problems and our approach efficiently avoids the deadlock for the Sokoban problem. We also show that bad effect that can be usually observed by introducing a priori knowledge into reinforcement learning process can be restrained by a method that decreases a rate of using knowledge during learning.