Prakash, M and Murty, Narasimha M (1995) A genetic approach for selection of (near-) optimal subsets of principal components for discrimination. In: Pattern Recognition Letters, 16 (8). pp. 781-787.
Restricted to Registered users only
Download (563Kb) | Request a copy
Principal Component Analysis (PCA) is being used both in the preprocessor to a feed-forward neural network and in the Subspace Pattern Recognition Method (SPRM). Most of the classifiers based on PCA use the first few Principal Components (PCs) associated with dominant eigenvalues of the pattern covariance matrix. Recent investigations reveal that considering PCs corresponding to the non-dominant eigenvalues will result in the design of better classifiers in certain application domains: the PCs that are most useful for discrimination may fall in the entire spectrum of PCs. Finding an optimal subset of PCs which maximizes the classification rate of a selected classifier is computationally expensive as the search space increases exponentially with the increase in either the dimensionality or the number of classes. In this paper, an approach based on genetic algorithms is proposed to search for an optimal subset of PCs. SPRM is used as the classifier because of its computational simplicity. The proposed approach is tested on two real data sets, and the results obtained are better than those obtained with Clafic, an algorithm used to choose the basis vectors in the SPRM. Although the PCs selected were from the entire spectrum, a considerable reduction in dimensionality was still present.
|Item Type:||Journal Article|
|Additional Information:||Copyright of this article belongs to Elsevier.|
|Department/Centre:||Division of Electrical Sciences > Computer Science & Automation (Formerly, School of Automation)|
|Date Deposited:||26 Mar 2007|
|Last Modified:||19 Sep 2010 04:35|
Actions (login required)