Convergence analysis of an augmented algorithm for fully complex-valued neural networks.

Abstract

This paper presents an augmented algorithm for fully complex-valued neural network based on Wirtinger calculus, which simplifies the derivation of the algorithm and eliminates the Schwarz symmetry restriction on the activation functions. A unified mean value theorem is first established for general functions of complex variables, covering the analytic functions, non-analytic functions and real-valued functions. Based on so introduced theorem, convergence results of the augmented algorithm are obtained under mild conditions. Simulations are provided to support the analysis.