Clustering quality based feature selection method
Machine Graphics & Vision International Journal
ECCV'06 Proceedings of the 9th European conference on Computer Vision - Volume Part II
Hi-index | 0.00 |
The problem of selecting a subset of relevant features in apotentially overwhelming quantity of data is classic and found inmany branches of science including - examples in computer vision,text processing and more recently bio-informatics are abundant. Inthis work we present a definition of '"relevancy" based on spectralproperties of the Affinity (or Laplacian) of the features'measurement matrix. The feature selection process is then based ona continuous ranking of the features defined by a least-squaresoptimization process. A remarkable property of the featurerelevance function is that sparse solutions for the ranking valuesnaturally emerge as a result of a "biased non-negativity" of a keymatrix in the process. As a result, a simple least-squaresoptimization process converges onto a sparse solution, i.e., aselection of a subset of features which form a local maxima overthe relevance function. The feature selection algorithm can beembedded in both unsupervised and supervised inference problems andempirical evidence show that the feature selections typicallyachieve high accuracy even when only a small fraction of thefeatures are relevant.