PCA Approach on Morphological Classification of Galaxies

The paper presents some results of the attempt on applying a PCA-based classifier for solving problem of morphological classification of galaxies according to the Hubble scheme. A variant of the PCA based classification method is adapted for solving the problem of galaxy image clustering. The performed tests pointed out that the first 24 principal components contain enough information to assure proper resizing for galaxy classification purposes. The skeletons composed from the first principal components for each class can be computed either by a first order approximation technique or, in case there are few principal components, using an exact method. The final section presents a series of conclusions and experimental results confirming the performance of the proposed algorithm, together with a comparative analysis against the perceptron algorithm.