An adaptive neural network A/D converter based on CMOS/memristor hybrid design

Abstract

A memristor is regarded as a promising device for modeling synapses in the realization of artificial neural systems for its nanoscale size, analog storage properties, low energy and non-volatility. In this letter, an adaptive T-Model neural network based on CMOS/memristor hybrid design is proposed to perform the analog-to-digital conversion without oscillations. The circuit is composed of CMOS neurons and memristor synapses. The A/D converter (ADC) is trained by the least mean square (LMS) algorithm. The conductance of the memristors can be adjusted to convert input voltages with different ranges, which makes the ADC flexible. Using memristors as synapses in neuromorphic circuits can potentially offer high density.