Title :
Principal-feature classification
Author :
Tufts, Donald W. ; Li, Qi
Author_Institution :
Dept. of Electr. & Comput. Eng., Rhode Island Univ., Kingston, RI, USA
fDate :
31 Aug-2 Sep 1995
Abstract :
The concept of classification using principal features is presented. The principal features defined in this paper are analogous to principal components in statistics and linear algebra. Neural network training can be done by sequential identification of principal features and corresponding pruning of the training data. Two neural network simplification algorithms, lossless and lossy simplifications, make the the classifier design more efficient. The design procedure is compared with other classifier design algorithms
Keywords :
learning (artificial intelligence); linear algebra; neural nets; pattern classification; signal detection; statistical analysis; linear algebra; lossless simplification; lossy simplifications; neural network; principal-feature network; sequential identification; signal recognition; statistics; Algebra; Algorithm design and analysis; Backpropagation algorithms; Neural networks; Performance analysis; Principal component analysis; Statistics; Testing; Training data; Vectors;
Conference_Titel :
Neural Networks for Signal Processing [1995] V. Proceedings of the 1995 IEEE Workshop
Conference_Location :
Cambridge, MA
Print_ISBN :
0-7803-2739-X
DOI :
10.1109/NNSP.1995.514886
