Performance analysis of q-valued dimensions Parametric Vector Neural Network classifier

Abstract

Neural network ensemble techniques have been shown to be very accurate classification techniques. However, in some real-life applications, a number of classifiers required to achieve a reasonable accuracy is enormously large and hence very space consuming. This paper introduces special neural method, Parametric Vector Neural Network (VNN), which has great associative memory and high performance. Parametric VNN analyzed using various size of database having randomly created patterns, noise levels, and fixed q-dimensions. The result shows that it has capacity much greater than conventional Neural Networks. Once T matrix is created for the stored patterns in Database, most similar pattern with the input one can be achieved easily by just multiplying two matrices. The resulting associative memory can recognize highly noisy and correlate input patterns.