HAL will be down for maintenance from Friday, June 10 at 4pm through Monday, June 13 at 9am. More information
Skip to Main content Skip to Navigation
Conference papers

MemSE: Fast MSE Prediction for Noisy Memristor-Based DNN Accelerators

Jonathan Kern 1, 2, 3 Sébastien Henwood 3 Gonçalo Mordido 3, 4 Elsa Dupraz 1, 5 Abdeldjalil Aissa El Bey 1, 2 Yvon Savaria 3 François Leduc-Primeau 3
2 Lab-STICC_COSYDE - Equipe Communication System Design
Lab-STICC - Laboratoire des sciences et techniques de l'information, de la communication et de la connaissance : UMR6285
5 Lab-STICC_CODES - Equipe CODES
Lab-STICC - Laboratoire des sciences et techniques de l'information, de la communication et de la connaissance : UMR6285
Abstract : Memristors enable the computation of matrix-vector multiplications (MVM) in memory and, therefore, show great potential in highly increasing the energy efficiency of deep neural network (DNN) inference accelerators. However, computations in memristors suffer from hardware non-idealities and are subject to different sources of noise that may negatively impact system performance. In this work, we theoretically analyze the mean squared error of DNNs that use memristor crossbars to compute MVM. We take into account both the quantization noise, due to the necessity of reducing the DNN model size, and the programming noise, stemming from the variability during the programming of the memristance value. Simulations on pretrained DNN models showcase the accuracy of the analytical prediction. Furthermore the proposed method is almost two order of magnitude faster than Monte-Carlo simulation, thus making it possible to optimize the implementation parameters to achieve minimal error for a given power constraint.
Document type :
Conference papers
Complete list of metadata

https://hal-imt.archives-ouvertes.fr/hal-03654471
Contributor : Abdeldjalil Aïssa-El-Bey Connect in order to contact the contributor
Submitted on : Thursday, April 28, 2022 - 4:10:41 PM
Last modification on : Wednesday, May 4, 2022 - 3:41:42 AM

File

 Restricted access
To satisfy the distribution rights of the publisher, the document is embargoed until : 2022-10-28

Please log in to resquest access to the document

Identifiers

  • HAL Id : hal-03654471, version 1

Citation

Jonathan Kern, Sébastien Henwood, Gonçalo Mordido, Elsa Dupraz, Abdeldjalil Aissa El Bey, et al.. MemSE: Fast MSE Prediction for Noisy Memristor-Based DNN Accelerators. IEEE international Conference on Artificial Intelligence Circuits and Systems (AICAS) 2022, Jun 2022, Incheon, South Korea. ⟨hal-03654471⟩

Share

Metrics

Record views

0