An Information-Theoretic Approach to Feature Extraction in Competitive Learning

R. Kamimura, T. Taniguchi, and R. Kitajima (Japan)


Mutual information, conditional information, competitive learning, selforganizing maps


In this paper, we propose a new information-theoretic ap proach to competitive learning and self-organizing maps. We use several information-theoretic measures such as con ditional information and information losses to extract main features in input patterns. For each competitive unit, condi tional information content is used to show how much infor mation on input patterns is contained. In addition, for de tecting the importance of each variable, information losses are introduced. The information loss is defined by differ ence between information with all input units and informa tion without an input unit. We applied the method to an artificial data, the Iris problem and a student survey. In all cases, experimental results showed that main features in in put patterns were clearly detected.

Important Links:

Go Back