Training the self-organizing feature map using hybrids of genetic and Kohonen methods

The self-organizing feature map is expected to produce a topologically correct mapping between input and output spaces. This mapping is usually found with the Kohonen learning rule which is sensitive to its parameter values. A poor choice of parameters results in a mapping that may not be topologically correct. In this paper, we describe a hybrid algorithm of genetic methods with Kohonen learning that avoids this problem. Experimental results show that this algorithm always results in a topologically correct mapping