Memristive Neural Networks Application In Predicting of Health Disorders

Publication Date

1-1-2023

Document Type

Conference Proceeding

Publication Title

Lecture Notes in Engineering and Computer Science

Volume

2245

First Page

94

Last Page

99

Abstract

This project focuses on designing a Memristive neural network (MNN) which proves to offer the same level of accuracy when compared with CMOS based DNN but consumes low power. The designed MNNs are simulated using modern CAD simulation frameworks for 22nm technology node and compared for the total leakage power with existing CMOS based neural network systems. The simulated MNNs estimate a leakage power of ~131 μW which is less compared to CMOS based neural network’s power and an area of ~10mm2 which is comparable with the size constraints for CMOS based implementations as studied.

Keywords

CMOS based implantation, Memristor, Neural networks

Department

Electrical Engineering

This document is currently not available here.

Share

COinS