Convergence Analysis of Complex Valued Multiplicative Neural Network for Various Activation Functions

In a complex valued neural network (CVNN) the weights, threshold, inputs and outputs are all complex numbers. Researchers have proposed many complex activation functions which can approximate a continuous complex valued function for CVNN node processing. The choice of an activation function determines the convergence of the complex back propagation algorithm and its generalization characteristics. In this paper we have compared the performance of various activation functions on the complex XOR problem for the complex multiplicative neural network.