Fault immunization for the central vector of a radial basis function neuron

Abstract

Radial basis function can be efficiently applied to the problems of pattern classification and machine learning and is also feasible to implement it on a VLSI chip. Typically, a radial basis function is usually realized by a Gaussian distribution function. Although, this function is popular and has some useful properties, rotating, translating, translating, and scaling the shape of this function in a high dimensional require a costly learning time. In this thesis, two related issues are considered. The first issue concerns the problem of developing a new radial basis function with less learning. The second issue focuses on the mathematical model of fault immunization of the proposed elliptic radial elliptic radial basis function. We consider only the data in a two dimensional space and test our solutions with some benchmarked data. The immunization technique increases the immunization degree ranging from 2% to 26%