A region selecting method which performs observation and action in the multi-resolution environment

Authors:
Toshihiro Matsui;Hiroshi Matsuo;Akira Iwata
Affiliations:
Department of Electrical & Computer Engineering, Nagoya Institute of Technology, Nagoya, Japan;Department of Electrical & Computer Engineering, Nagoya Institute of Technology, Nagoya, Japan;Department of Electrical & Computer Engineering, Nagoya Institute of Technology, Nagoya, Japan
Venue:
PRICAI'00 Proceedings of the 6th Pacific Rim international conference on Artificial intelligence
Year:
2000

Citing 3
Cited 0

k-Certainty exploration method: an action selector to identify the environment in reinforcement learning

Artificial Intelligence - Special issue: artificial intelligence research in Japan
A dynamic and multiresolution model of visual attention and its application to facial landmark detection

Computer Vision and Image Understanding
Reinforcement learning: a survey

Journal of Artificial Intelligence Research

Quantified Score

Hi-index	0.00

Visualization

Abstract

We propose a method for selecting the characteristic region in the environment based on the occurrence probability of the pattern. If the occurrence probability of the pattern is unknown in initial stage, estimation of the distribution of the pattern and selection of the characteristic region must be done simultaneously. We noticed that a method for exploration of the state-space in reinforcement learning was similar to such task. Then, we propose a method for selecting the characteristic region by repeating observation and action in the environment. In the observation using only one resolution, the position in the environment can not be decided. The multi-resolution concept is introduced in order to solve this problem. The experimental result shows that the characteristic region is selected from the environment.