A signal separation view of classification
This offers a novel theoretical framework for classification that could improve efficiency in domains like hyperspectral imaging and document analysis, though it appears incremental as it adapts existing signal processing techniques.
The paper tackles classification by reframing it as a signal separation problem, using localized trigonometric polynomial kernels to separate class distributions and determine the number of classes with minimal labeled data, achieving perfect classification in theory and demonstrating results on simulated and real datasets like Salinas and Indian Pines.
The problem of classification in machine learning has often been approached in terms of function approximation. In this paper, we propose an alternative approach for classification in arbitrary compact metric spaces which, in theory, yields both the number of classes, and a perfect classification using a minimal number of queried labels. Our approach uses localized trigonometric polynomial kernels initially developed for the point source signal separation problem in signal processing. Rather than point sources, we argue that the various classes come from different probability distributions. The localized kernel technique developed for separating point sources is then shown to separate the supports of these distributions. This is done in a hierarchical manner in our MASC algorithm to accommodate touching/overlapping class boundaries. We illustrate our theory on several simulated and real life datasets, including the Salinas and Indian Pines hyperspectral datasets and a document dataset.