Advances in Self-Organizing Maps and Learning Vector Quantization

Abstract

Next-generation sequencing techniques produce an enormous amount of sequence data. Analyzing these sequences requires an efficient method that can handle large amounts of data. Self-organizing maps (SOMs), which use the frequencies of N-tuples, can categorize sets of DNA sequences with unsupervised learning. In this study, SOM using correlation coefficients among nucleotides was proposed, and its performance was examined in the experiments through mapping experiments of the genome sequences of several species and classification experiments using Pareto learning SOMs. © Springer International Publishing Switzerland 2014.