Ordered versus pre and post process vectors in the supervised methods of A.N.N

Abstract EN

This paper proposes some useful modifications to the data before and after applying the back-propagation learning algorithm, to overcome the problem of learning ordered data.

And so considers a Multilayer feed-forward neural networks (MFNN) with momentum, and Counter propagation (CP) which is a combined classifier uses the generalization capabilities of both the Back-propagation (BP) and Learning Vector Quantization (LVQ), the proposed method applied on both some mathematical functions and image filtering, the results compared with standard MFNN and CP.