Fuzzy-SOMを用いた自律行動体の行動様式の類別化

抄録

A method based on Self-Organizing feature Map (SOM) for classifying behavior patterns of autonomous actors has been proposed. This paper shows that the method has been able to display a classified behavior of gold fish after a shock was applied. The pattern of behavior is calculated using five-dimensional input vectors formed from animated pictures. This classification method is improved so that the displaying map may not be influenced by initial conditions. In this case, the Euclidean distance on competitive layer is arranged by the fuzzy inference. The Fuzzy-SOM is useful to analyze acts of the athlete in sports and refugees in an accident, because this method can deal with animated pictures. An application to a soccer game is also presented.