A multi-attribute classification method to solve the problem of dimensionality

Abstract

Classification is one of the most important areas of machine learning. However, there are numerous applications where the quantity of attributes is very large, rendering the usage of conventional classifiers very slow or even impossible. The classifier method in this paper is proposed for such problems. Using the assumption that very large problem spaces are typically sparse as well (considering the stored knowledge), it maps the multi-dimensional problem space into a sequential combination of two-dimensional subdomains. The classifier is easy to implement, fast, and capable of recognizing patterns that are similar to known ones.