publication . Other literature type . Conference object . 2015 . Embargo end date: 10 Nov 2015

Practical gradient-descent for memristive crossbars

Name: Practical gradient-descent for memristive crossbars
Keywords: Institute of Neuroinformatics, 570 Life sciences; biology

Piotr Dudek; Manu V Nair;

Open Access

Published: 10 Nov 2015

Publisher: 2015 International Conference on Memristive Systems (MEMRISYS)

Country: Switzerland

Summary

Abstract

This paper discusses implementations of gradientdescent based learning algorithms on memristive crossbar arrays. The Unregulated Step Descent (USD) is described as a practical algorithm for feed-forward on-line training of large crossbar arrays. It allows fast feed-forward fully parallel on-line hardware based learning, without requiring accurate models of the memristor behaviour and precise control of the programming pulses. The effect of device parameters, training parameters, and device variability on the learning performance of crossbar arrays trained using the USD algorithm has been studied via simulations.

doi: 10.5167/uzh-132662 , 10.1109/memrisys.2015.7378392

Subjects

free text keywords: Institute of Neuroinformatics, 570 Life sciences; biology, Computer science, Practical algorithm, Convergence (routing), Computer engineering, Crossbar switch, Implementation, Memristor, law.invention, law, Device parameters, Electronic engineering, Gradient descent

Related Organizations