Neural-net method for dual subspace pattern recognition
L. Xu, Adam Krzyżak, Erkki Oja · 2002
A novel version of the subspace pattern recognition method, called the dual subspace pattern recognition (DSPR) method, is proposed, and a neural-network model with a combination of modified Hebbian and anti-Hebbian learning rules is developed for implementing the DSPR method. An experimental comparison was made on an example data set by using this model and a three-layer forward net with back propagation learning. The results demonstrate that this model can outperform the back propagation model in some applications.>