An Improved ISOMAP for Classification

摘要

ISOMAP behaves in an unsupervised manner and therefore works less effectively for classification. In this paper, an improved ISOMAP for Classification task, called ISOMAP-C, is proposed, which employs label information to guide the dimensionality reduction. Firstly, within-class neighborhood graphs are constructed over each sub dataset of the same class according to label information. Secondly, it searches for the between-class adjacent edges with the shortest distance, which is multiplied by scaling factor greater than one so that low dimensional data set after mapping become more compact within class and more separate between classes. Finally, the mapping function from original high dimensional space to low dimensional space can be approximately modeled using Back-Propagation neural network combined with genetic algorithm. Experimental results show that ISOMAP-C is effective. This article is designed to help in the contribution for the Journal of Information and Computational Science.