Feedforward multilayer phase-based neural networks

Abstract

Complex-Valued Neural Networks (CVNNs) are Artificial Neural Networks (ANNs) which function using complex numbers - they have complex-valued parameters and accept complex-valued inputs. Phase-Based Neurons (PBNs) are simple CVNNs that use for the internal weights complex numbers with the modulus 1, the only adaptable parameters being the phases of the weights. We present in this paper some limitations of the Continuous Phase-Based Neuron (CPBN) and describe the structure of a Feedforward Multilayer Phase-Based Neural Network (MLPBN) and its training using an adaptation of the backpropagation algorithm.