Quantization of Deep Neural Networks for Accumulator-constrained Processors

We introduce an Artificial Neural Network (ANN) quantization methodology for platforms without wide accumulation registers. This enables fixed-point model deployment on embedded compute platforms that are not specifically designed for large kernel computations (i.e. accumulator-constrained processor...

Full description

Saved in:
Bibliographic Details
Published in:arXiv.org 2020-04
Main Authors: de Bruin, Barry, Zivkovic, Zoran, Corporaal, Henk
Format: Article
Language:English
Subjects:
Accumulators
Artificial neural networks
Floating point arithmetic
Image classification
Measurement
Microprocessors
Model accuracy
Neural networks
Optimization
Platforms
Processors
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
cited_by
cites
container_end_page
container_issue
container_start_page
container_title arXiv.org
container_volume
creator de Bruin, Barry
Zivkovic, Zoran
Corporaal, Henk
description We introduce an Artificial Neural Network (ANN) quantization methodology for platforms without wide accumulation registers. This enables fixed-point model deployment on embedded compute platforms that are not specifically designed for large kernel computations (i.e. accumulator-constrained processors). We formulate the quantization problem as a function of accumulator size, and aim to maximize the model accuracy by maximizing bit width of input data and weights. To reduce the number of configurations to consider, only solutions that fully utilize the available accumulator bits are being tested. We demonstrate that 16-bit accumulators are able to obtain a classification accuracy within 1\% of the floating-point baselines on the CIFAR-10 and ILSVRC2012 image classification benchmarks. Additionally, a near-optimal \(2\times\) speedup is obtained on an ARM processor, by exploiting 16-bit accumulators for image classification on the All-CNN-C and AlexNet networks.
doi_str_mv 10.48550/arxiv.2004.11783
format article
fullrecord <record><control><sourceid>proquest</sourceid><recordid>TN_cdi_proquest_journals_2395072454</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2395072454</sourcerecordid><originalsourceid>FETCH-LOGICAL-a524-e6c63c7a36c84d126aacafd28f67607b66b4ddde91b0ed1d812202fb9ca9079c3</originalsourceid><addsrcrecordid>eNotzUlLAzEYgOEgCJbaH-At4HnG5Ms2cyx1xeICvZdvssDUOqlZVPz1FvT03N6XkAvOWtkpxa4wfY-fLTAmW85NJ07IDITgTScBzsgi5x1jDLQBpcSMPL5WnMr4g2WME42BXnt_oE--JtwfKV8xvWUaYqJLa-t73WOJqbFxyiXhOHlHX1K0PueY8jk5DbjPfvHvnGxubzar-2b9fPewWq4bVCAbr60W1qDQtpOOg0a0GBx0QRvNzKD1IJ1zvucD8467jgMwCENvsWemt2JOLv-yhxQ_qs9lu4s1TcfjFkSvmAGppPgF5DRQGA</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2395072454</pqid></control><display><type>article</type><title>Quantization of Deep Neural Networks for Accumulator-constrained Processors</title><source>Publicly Available Content Database</source><creator>de Bruin, Barry ; Zivkovic, Zoran ; Corporaal, Henk</creator><creatorcontrib>de Bruin, Barry ; Zivkovic, Zoran ; Corporaal, Henk</creatorcontrib><description>We introduce an Artificial Neural Network (ANN) quantization methodology for platforms without wide accumulation registers. This enables fixed-point model deployment on embedded compute platforms that are not specifically designed for large kernel computations (i.e. accumulator-constrained processors). We formulate the quantization problem as a function of accumulator size, and aim to maximize the model accuracy by maximizing bit width of input data and weights. To reduce the number of configurations to consider, only solutions that fully utilize the available accumulator bits are being tested. We demonstrate that 16-bit accumulators are able to obtain a classification accuracy within 1\% of the floating-point baselines on the CIFAR-10 and ILSVRC2012 image classification benchmarks. Additionally, a near-optimal \(2\times\) speedup is obtained on an ARM processor, by exploiting 16-bit accumulators for image classification on the All-CNN-C and AlexNet networks.</description><identifier>EISSN: 2331-8422</identifier><identifier>DOI: 10.48550/arxiv.2004.11783</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Accumulators ; Artificial neural networks ; Floating point arithmetic ; Image classification ; Measurement ; Microprocessors ; Model accuracy ; Neural networks ; Optimization ; Platforms ; Processors</subject><ispartof>arXiv.org, 2020-04</ispartof><rights>2020. This work is published under http://arxiv.org/licenses/nonexclusive-distrib/1.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://www.proquest.com/docview/2395072454?pq-origsite=primo$$EHTML$$P50$$Gproquest$$Hfree_for_read</linktohtml><link.rule.ids>777,781,25735,27907,36994,44572</link.rule.ids></links><search><creatorcontrib>de Bruin, Barry</creatorcontrib><creatorcontrib>Zivkovic, Zoran</creatorcontrib><creatorcontrib>Corporaal, Henk</creatorcontrib><title>Quantization of Deep Neural Networks for Accumulator-constrained Processors</title><title>arXiv.org</title><description>We introduce an Artificial Neural Network (ANN) quantization methodology for platforms without wide accumulation registers. This enables fixed-point model deployment on embedded compute platforms that are not specifically designed for large kernel computations (i.e. accumulator-constrained processors). We formulate the quantization problem as a function of accumulator size, and aim to maximize the model accuracy by maximizing bit width of input data and weights. To reduce the number of configurations to consider, only solutions that fully utilize the available accumulator bits are being tested. We demonstrate that 16-bit accumulators are able to obtain a classification accuracy within 1\% of the floating-point baselines on the CIFAR-10 and ILSVRC2012 image classification benchmarks. Additionally, a near-optimal \(2\times\) speedup is obtained on an ARM processor, by exploiting 16-bit accumulators for image classification on the All-CNN-C and AlexNet networks.</description><subject>Accumulators</subject><subject>Artificial neural networks</subject><subject>Floating point arithmetic</subject><subject>Image classification</subject><subject>Measurement</subject><subject>Microprocessors</subject><subject>Model accuracy</subject><subject>Neural networks</subject><subject>Optimization</subject><subject>Platforms</subject><subject>Processors</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2020</creationdate><recordtype>article</recordtype><sourceid>PIMPY</sourceid><recordid>eNotzUlLAzEYgOEgCJbaH-At4HnG5Ms2cyx1xeICvZdvssDUOqlZVPz1FvT03N6XkAvOWtkpxa4wfY-fLTAmW85NJ07IDITgTScBzsgi5x1jDLQBpcSMPL5WnMr4g2WME42BXnt_oE--JtwfKV8xvWUaYqJLa-t73WOJqbFxyiXhOHlHX1K0PueY8jk5DbjPfvHvnGxubzar-2b9fPewWq4bVCAbr60W1qDQtpOOg0a0GBx0QRvNzKD1IJ1zvucD8467jgMwCENvsWemt2JOLv-yhxQ_qs9lu4s1TcfjFkSvmAGppPgF5DRQGA</recordid><startdate>20200424</startdate><enddate>20200424</enddate><creator>de Bruin, Barry</creator><creator>Zivkovic, Zoran</creator><creator>Corporaal, Henk</creator><general>Cornell University Library, arXiv.org</general><scope>8FE</scope><scope>8FG</scope><scope>ABJCF</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BGLVJ</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>HCIFZ</scope><scope>L6V</scope><scope>M7S</scope><scope>PIMPY</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>PRINS</scope><scope>PTHSS</scope></search><sort><creationdate>20200424</creationdate><title>Quantization of Deep Neural Networks for Accumulator-constrained Processors</title><author>de Bruin, Barry ; Zivkovic, Zoran ; Corporaal, Henk</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a524-e6c63c7a36c84d126aacafd28f67607b66b4ddde91b0ed1d812202fb9ca9079c3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2020</creationdate><topic>Accumulators</topic><topic>Artificial neural networks</topic><topic>Floating point arithmetic</topic><topic>Image classification</topic><topic>Measurement</topic><topic>Microprocessors</topic><topic>Model accuracy</topic><topic>Neural networks</topic><topic>Optimization</topic><topic>Platforms</topic><topic>Processors</topic><toplevel>online_resources</toplevel><creatorcontrib>de Bruin, Barry</creatorcontrib><creatorcontrib>Zivkovic, Zoran</creatorcontrib><creatorcontrib>Corporaal, Henk</creatorcontrib><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>Materials Science &amp; Engineering Collection</collection><collection>ProQuest Central (Alumni Edition)</collection><collection>ProQuest Central UK/Ireland</collection><collection>ProQuest Central Essentials</collection><collection>ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central Korea</collection><collection>SciTech Premium Collection</collection><collection>ProQuest Engineering Collection</collection><collection>Engineering Database</collection><collection>Publicly Available Content Database</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><collection>ProQuest Central China</collection><collection>Engineering Collection</collection><jtitle>arXiv.org</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>de Bruin, Barry</au><au>Zivkovic, Zoran</au><au>Corporaal, Henk</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Quantization of Deep Neural Networks for Accumulator-constrained Processors</atitle><jtitle>arXiv.org</jtitle><date>2020-04-24</date><risdate>2020</risdate><eissn>2331-8422</eissn><abstract>We introduce an Artificial Neural Network (ANN) quantization methodology for platforms without wide accumulation registers. This enables fixed-point model deployment on embedded compute platforms that are not specifically designed for large kernel computations (i.e. accumulator-constrained processors). We formulate the quantization problem as a function of accumulator size, and aim to maximize the model accuracy by maximizing bit width of input data and weights. To reduce the number of configurations to consider, only solutions that fully utilize the available accumulator bits are being tested. We demonstrate that 16-bit accumulators are able to obtain a classification accuracy within 1\% of the floating-point baselines on the CIFAR-10 and ILSVRC2012 image classification benchmarks. Additionally, a near-optimal \(2\times\) speedup is obtained on an ARM processor, by exploiting 16-bit accumulators for image classification on the All-CNN-C and AlexNet networks.</abstract><cop>Ithaca</cop><pub>Cornell University Library, arXiv.org</pub><doi>10.48550/arxiv.2004.11783</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier EISSN: 2331-8422
ispartof arXiv.org, 2020-04
issn 2331-8422
language eng
recordid cdi_proquest_journals_2395072454
source Publicly Available Content Database
subjects Accumulators
Artificial neural networks
Floating point arithmetic
Image classification
Measurement
Microprocessors
Model accuracy
Neural networks
Optimization
Platforms
Processors
title Quantization of Deep Neural Networks for Accumulator-constrained Processors
url http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-17T08%3A23%3A03IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Quantization%20of%20Deep%20Neural%20Networks%20for%20Accumulator-constrained%20Processors&rft.jtitle=arXiv.org&rft.au=de%20Bruin,%20Barry&rft.date=2020-04-24&rft.eissn=2331-8422&rft_id=info:doi/10.48550/arxiv.2004.11783&rft_dat=%3Cproquest%3E2395072454%3C/proquest%3E%3Cgrp_id%3Ecdi_FETCH-LOGICAL-a524-e6c63c7a36c84d126aacafd28f67607b66b4ddde91b0ed1d812202fb9ca9079c3%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_pqid=2395072454&rft_id=info:pmid/&rfr_iscdi=true