Open AccessLEARNING OF BIOLOGICAL BEHAVIOUR BY CLASSIFIER
Author(s) -
Hideaki Kanoh,
Akihiro Hosokawa
Publication year - 2014
Publication title -
computing
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.184
H-Index - 11
eISSN - 2312-5381
pISSN - 1727-6209
DOI - 10.47839/ijc.1.1.80
Subject(s) - classifier (uml) , computer science , artificial intelligence , machine learning , pattern recognition (psychology)
The objective of this paper is to investigate biological features of a virtual creature on the computer. It is assumed that the creature does not have a complex judgment, a sophisticated detector and has simple basic actions. We call it a bug. As a lower animal, an insect does not have complex brain, while it is seemed to be able to do complicated works. Specifically complex behaviour of an insect is considered to be based on a reflex action for the external stimuli. The classifier system is used for the generation of bug's behaviour rules in a complicated environment made in the computer. The classifier system is used for the generation of bug's behaviour rules in a complicated environment made in the computer. A classifier system is a reflex system, which generates an action soon after receiving a stimulus from the environment. This is quite similar to the feature of lower animals function of processing the information. The bug is released at some spot in the virtual world where baits are randomly located and there are obstacles and enemies. The classifier is rewarded only when the bug succeeds to obtain the bait. The more the bug learned, the more new classifier rules emerge: such as a rule to take a step forward or another rule to change the direction towards bait. At this stage, the bug learns the way to obtain the bait avoiding the obstacles and enemies by itself. The results of the observation of simulation experiments show that the chain of the continuous action which consists of multiple classifiers is arising.