A programmable on-chip BP learning neural network with enhanced neuron characteristics

Abstract

A circuit system of programmable on-chip BP (Back-Propagation) learning neural network with enhanced neuron characteristics is designed. The whole system comprises feedforward network, error back-propagation network and weight updating circuit. It has the merits of simplicity, programmability, speediness, low power consumption and high density. A novel neuron circuit with programmable parameters is proposed. It generates not only the sigmoidal function but also its derivative. HSPICE simulations are carried out on the neuron circuit using level 47 transistor models for a standard 1.2 /spl mu/m CMOS process. The results show that both functions are matched with their ideal functions very accurately. The non-linear partition and function fitness hardware simulations are carried out for the whole system. Both experiments verify the superior performance of this BP neural network with on-chip learning.