Cargando…

EnforceSNN: Enabling resilient and energy-efficient spiking neural network inference considering approximate DRAMs for embedded systems

Spiking Neural Networks (SNNs) have shown capabilities of achieving high accuracy under unsupervised settings and low operational power/energy due to their bio-plausible computations. Previous studies identified that DRAM-based off-chip memory accesses dominate the energy consumption of SNN processi...

Descripción completa

Detalles Bibliográficos
Autores principales: Putra, Rachmad Vidya Wicaksana, Hanif, Muhammad Abdullah, Shafique, Muhammad
Formato: Online Artículo Texto
Lenguaje:English
Publicado: Frontiers Media S.A. 2022
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9399768/
https://www.ncbi.nlm.nih.gov/pubmed/36033624
http://dx.doi.org/10.3389/fnins.2022.937782
_version_ 1784772600367415296
author Putra, Rachmad Vidya Wicaksana
Hanif, Muhammad Abdullah
Shafique, Muhammad
author_facet Putra, Rachmad Vidya Wicaksana
Hanif, Muhammad Abdullah
Shafique, Muhammad
author_sort Putra, Rachmad Vidya Wicaksana
collection PubMed
description Spiking Neural Networks (SNNs) have shown capabilities of achieving high accuracy under unsupervised settings and low operational power/energy due to their bio-plausible computations. Previous studies identified that DRAM-based off-chip memory accesses dominate the energy consumption of SNN processing. However, state-of-the-art works do not optimize the DRAM energy-per-access, thereby hindering the SNN-based systems from achieving further energy efficiency gains. To substantially reduce the DRAM energy-per-access, an effective solution is to decrease the DRAM supply voltage, but it may lead to errors in DRAM cells (i.e., so-called approximate DRAM). Toward this, we propose EnforceSNN, a novel design framework that provides a solution for resilient and energy-efficient SNN inference using reduced-voltage DRAM for embedded systems. The key mechanisms of our EnforceSNN are: (1) employing quantized weights to reduce the DRAM access energy; (2) devising an efficient DRAM mapping policy to minimize the DRAM energy-per-access; (3) analyzing the SNN error tolerance to understand its accuracy profile considering different bit error rate (BER) values; (4) leveraging the information for developing an efficient fault-aware training (FAT) that considers different BER values and bit error locations in DRAM to improve the SNN error tolerance; and (5) developing an algorithm to select the SNN model that offers good trade-offs among accuracy, memory, and energy consumption. The experimental results show that our EnforceSNN maintains the accuracy (i.e., no accuracy loss for BER ≤ 10(−3)) as compared to the baseline SNN with accurate DRAM while achieving up to 84.9% of DRAM energy saving and up to 4.1x speed-up of DRAM data throughput across different network sizes.
format Online
Article
Text
id pubmed-9399768
institution National Center for Biotechnology Information
language English
publishDate 2022
publisher Frontiers Media S.A.
record_format MEDLINE/PubMed
spelling pubmed-93997682022-08-25 EnforceSNN: Enabling resilient and energy-efficient spiking neural network inference considering approximate DRAMs for embedded systems Putra, Rachmad Vidya Wicaksana Hanif, Muhammad Abdullah Shafique, Muhammad Front Neurosci Neuroscience Spiking Neural Networks (SNNs) have shown capabilities of achieving high accuracy under unsupervised settings and low operational power/energy due to their bio-plausible computations. Previous studies identified that DRAM-based off-chip memory accesses dominate the energy consumption of SNN processing. However, state-of-the-art works do not optimize the DRAM energy-per-access, thereby hindering the SNN-based systems from achieving further energy efficiency gains. To substantially reduce the DRAM energy-per-access, an effective solution is to decrease the DRAM supply voltage, but it may lead to errors in DRAM cells (i.e., so-called approximate DRAM). Toward this, we propose EnforceSNN, a novel design framework that provides a solution for resilient and energy-efficient SNN inference using reduced-voltage DRAM for embedded systems. The key mechanisms of our EnforceSNN are: (1) employing quantized weights to reduce the DRAM access energy; (2) devising an efficient DRAM mapping policy to minimize the DRAM energy-per-access; (3) analyzing the SNN error tolerance to understand its accuracy profile considering different bit error rate (BER) values; (4) leveraging the information for developing an efficient fault-aware training (FAT) that considers different BER values and bit error locations in DRAM to improve the SNN error tolerance; and (5) developing an algorithm to select the SNN model that offers good trade-offs among accuracy, memory, and energy consumption. The experimental results show that our EnforceSNN maintains the accuracy (i.e., no accuracy loss for BER ≤ 10(−3)) as compared to the baseline SNN with accurate DRAM while achieving up to 84.9% of DRAM energy saving and up to 4.1x speed-up of DRAM data throughput across different network sizes. Frontiers Media S.A. 2022-08-10 /pmc/articles/PMC9399768/ /pubmed/36033624 http://dx.doi.org/10.3389/fnins.2022.937782 Text en Copyright © 2022 Putra, Hanif and Shafique. https://creativecommons.org/licenses/by/4.0/This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
spellingShingle Neuroscience
Putra, Rachmad Vidya Wicaksana
Hanif, Muhammad Abdullah
Shafique, Muhammad
EnforceSNN: Enabling resilient and energy-efficient spiking neural network inference considering approximate DRAMs for embedded systems
title EnforceSNN: Enabling resilient and energy-efficient spiking neural network inference considering approximate DRAMs for embedded systems
title_full EnforceSNN: Enabling resilient and energy-efficient spiking neural network inference considering approximate DRAMs for embedded systems
title_fullStr EnforceSNN: Enabling resilient and energy-efficient spiking neural network inference considering approximate DRAMs for embedded systems
title_full_unstemmed EnforceSNN: Enabling resilient and energy-efficient spiking neural network inference considering approximate DRAMs for embedded systems
title_short EnforceSNN: Enabling resilient and energy-efficient spiking neural network inference considering approximate DRAMs for embedded systems
title_sort enforcesnn: enabling resilient and energy-efficient spiking neural network inference considering approximate drams for embedded systems
topic Neuroscience
url https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9399768/
https://www.ncbi.nlm.nih.gov/pubmed/36033624
http://dx.doi.org/10.3389/fnins.2022.937782
work_keys_str_mv AT putrarachmadvidyawicaksana enforcesnnenablingresilientandenergyefficientspikingneuralnetworkinferenceconsideringapproximatedramsforembeddedsystems
AT hanifmuhammadabdullah enforcesnnenablingresilientandenergyefficientspikingneuralnetworkinferenceconsideringapproximatedramsforembeddedsystems
AT shafiquemuhammad enforcesnnenablingresilientandenergyefficientspikingneuralnetworkinferenceconsideringapproximatedramsforembeddedsystems