Search. Read. Cite.

Easy to search. Easy to read. Easy to cite with credible sources.

Information Technology Journal

Year: 2003  |  Volume: 2  |  Issue: 1  |  Page No.: 40 - 43

Feature Extraction by Using Non-linear and Unsupervised Neural Networks

A. Jalil, I.M. Qureshi, A. Naveed and T.a. Cheema


Feature extraction is fairly popular in pattern recognition and classification of images. In this paper we propose an unsupervised learning algorithm for neural networks that are used in feature extraction problem. These learning algorithms use genetic algorithm as a searching technique for global minimum of error performance surface and LMS algorithm for final convergence to the global minimum. These learning algorithms used Sammon`s stress as criterion for getting feature with maximum inter pattern distances and minimum intra pattern distances. A common attribute of these learning algorithms is that they are adaptive in nature, which makes them suitable in environments when the distribution of patterns in the feature space changes with respect to time.

Cited References Fulltext