Semi-Supervised Neural Classifier using Memristive Nanodevices

Ahmad M. Sheri, Seungjong No, and Moongu Jeon

Keywords

Robotics, Biomedical Computing, Nanowire Memristors, SOM

Abstract

Nanowire memristor devices seem a natural electronic substitute of a synaptic junction in neuromorphic devices. The paper explores one of the architectures of neural computation which classifies an input set into distinct classes. A specific case of character recognition is implemented using neural networks which use nanowire memristive edges and CMOS based neural nodes. Simulation were conducted to classify 10 digits into distinct classes. The network is then tested under noisy conditions and shows promising results.

Important Links:



Go Back