CNNs with radial basis input function

This paper proposes a Cellular Neural Network (CNN) [1] model with radial basis input function (henceforth called as radial basis input CNN) for improving function approximation ability of CNNs. The model can be viewed as a cascade of two units: First unit is a multi-input, multi-output Radial Basis Function Network (RBFN) [2], second unit is original CNN model. The weights and centers of RBFN unit are chosen identical for all RBFN outputs yielding a space-invariant connection weight pattern over the network. With such a weight sharing property, the proposed model becomes a special kind of nonlinear B-template CNN [3]. The ability of the radial basis input CNN model in approximation to functions as its input-(steady state)output mapping is examined on edge detection task for noisy images. Herein, a modified version of Recurrent Perceptron Learning Algorithm (RPLA) [4] is used for training radial basis input CNN.