Practical Gradient-Descent for Memristive Crossbars

M.V. Nair; P. Dudek

doi:10.1109/MEMRISYS.2015.7378392

Practical Gradient-Descent for Memristive Crossbars

Research output: Chapter in Book/Conference proceeding › Conference contribution › peer-review

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.

Original language	English
Title of host publication	International Conference on Memristive Systems, MEMRISYS 2015
Publisher	IEEE
ISBN (Electronic)	978-1-4673-9209-9
DOIs	https://doi.org/10.1109/MEMRISYS.2015.7378392
Publication status	Published - Nov 2015
Event	International Conference on Memristive Systems, MEMRISYS 2015 - Duration: 1 Jan 1824 → …

Conference

Conference	International Conference on Memristive Systems, MEMRISYS 2015
Period	1/01/24 → …

Access to Document

10.1109/MEMRISYS.2015.7378392

Cite this

@inproceedings{3a8f17c966c5400599edabe551199fad,

title = "Practical Gradient-Descent for Memristive Crossbars",

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.",

author = "M.V. Nair and P. Dudek",

year = "2015",

month = nov,

doi = "10.1109/MEMRISYS.2015.7378392",

language = "English",

booktitle = "International Conference on Memristive Systems, MEMRISYS 2015",

publisher = "IEEE",

address = "United States",

note = "International Conference on Memristive Systems, MEMRISYS 2015 ; Conference date: 01-01-1824",

}

TY - GEN

T1 - Practical Gradient-Descent for Memristive Crossbars

AU - Nair, M.V.

AU - Dudek, P.

PY - 2015/11

Y1 - 2015/11

N2 - 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.

AB - 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.

U2 - 10.1109/MEMRISYS.2015.7378392

DO - 10.1109/MEMRISYS.2015.7378392

M3 - Conference contribution

BT - International Conference on Memristive Systems, MEMRISYS 2015

PB - IEEE

T2 - International Conference on Memristive Systems, MEMRISYS 2015

Y2 - 1 January 1824

ER -

Practical Gradient-Descent for Memristive Crossbars

Abstract

Conference

Access to Document

Fingerprint

Cite this