Information-theoretic Self-organizing Maps and Their Application to Road and Chemical Compound Classification

R. Kamimura and H. Takeuchi (Japan)


: mutual information maximization, competitive learning, self-organizing maps, chemical compound,road classification


In this paper, we apply the cost-sensitive self-organizing maps to road and chemical compound classification. The cost-sensitive self-organizing maps aim to produce equiprobabilistic topographic maps by maximizing infor mation. Experimental results confirmed that cooperation processes can increase significantly information content in input patterns. With entropy maximization, cooperation processes can produce clearer topographic maps with lit tle conflict neurons.

Important Links:

Go Back