L2PF -- Learning to Prune Faster

L2PF -- Learning to Prune Faster

Various applications in the field of autonomous driving are based on convolutional neural networks (CNNs), especially for processing camera data. The optimization of such CNNs is a major challenge in continuous development. Newly learned features must be brought into vehicles as quickly as possible,...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Vemparala, Manoj-Rohit, Fasfous, Nael, Frickenstein, Alexander, Moraly, Mhd Ali, Jamal, Aquib, Frickenstein, Lukas, Unger, Christian, Nagaraja, Naveen-Shankar, Stechele, Walter
Format: Artikel
Sprache:eng
Schlagworte:
Computer Science - Artificial Intelligence
Computer Science - Computer Vision and Pattern Recognition
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Vemparala, Manoj-Rohit
Fasfous, Nael
Frickenstein, Alexander
Moraly, Mhd Ali
Jamal, Aquib
Frickenstein, Lukas
Unger, Christian
Nagaraja, Naveen-Shankar
Stechele, Walter
description Various applications in the field of autonomous driving are based on convolutional neural networks (CNNs), especially for processing camera data. The optimization of such CNNs is a major challenge in continuous development. Newly learned features must be brought into vehicles as quickly as possible, and as such, it is not feasible to spend redundant GPU hours during compression. In this context, we present Learning to Prune Faster which details a multi-task, try-and-learn method, discretely learning redundant filters of the CNN and a continuous action of how long the layers have to be fine-tuned. This allows us to significantly speed up the convergence process of learning how to find an embedded-friendly filter-wise pruned CNN. For ResNet20, we have achieved a compression ratio of 3.84 x with minimal accuracy degradation. Compared to the state-of-the-art pruning method, we reduced the GPU hours by 1.71 x.
doi_str_mv 10.48550/arxiv.2101.02663
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2101_02663</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2101_02663</sourcerecordid><originalsourceid>FETCH-LOGICAL-a673-8525e6f113d8d8698fc0c37e420af7f8e029e91e1faafd16427afe05439038ea3</originalsourceid><addsrcrecordid>eNotzrEOgjAUQNEuDgb9ACf7A8XXlpYyGiJqQqKDO3mBV0OiaAoa_XsjOt3t5jC2kBAnzhhYYXi1z1hJkDEoa_WU8VIdCy4ELwlD13ZnPtz4MTw64gX2A4UZm3i89DT_N2KnYnPKd6I8bPf5uhRoUy2cUYasl1I3rnE2c76GWqeUKECfekegMsokSY_oG2kTlaInMInOQDtCHbHlbzsSq3torxje1ZdajVT9AY6qNjE</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>L2PF -- Learning to Prune Faster</title><source>arXiv.org</source><creator>Vemparala, Manoj-Rohit ; Fasfous, Nael ; Frickenstein, Alexander ; Moraly, Mhd Ali ; Jamal, Aquib ; Frickenstein, Lukas ; Unger, Christian ; Nagaraja, Naveen-Shankar ; Stechele, Walter</creator><creatorcontrib>Vemparala, Manoj-Rohit ; Fasfous, Nael ; Frickenstein, Alexander ; Moraly, Mhd Ali ; Jamal, Aquib ; Frickenstein, Lukas ; Unger, Christian ; Nagaraja, Naveen-Shankar ; Stechele, Walter</creatorcontrib><description>Various applications in the field of autonomous driving are based on convolutional neural networks (CNNs), especially for processing camera data. The optimization of such CNNs is a major challenge in continuous development. Newly learned features must be brought into vehicles as quickly as possible, and as such, it is not feasible to spend redundant GPU hours during compression. In this context, we present Learning to Prune Faster which details a multi-task, try-and-learn method, discretely learning redundant filters of the CNN and a continuous action of how long the layers have to be fine-tuned. This allows us to significantly speed up the convergence process of learning how to find an embedded-friendly filter-wise pruned CNN. For ResNet20, we have achieved a compression ratio of 3.84 x with minimal accuracy degradation. Compared to the state-of-the-art pruning method, we reduced the GPU hours by 1.71 x.</description><identifier>DOI: 10.48550/arxiv.2101.02663</identifier><language>eng</language><subject>Computer Science - Artificial Intelligence ; Computer Science - Computer Vision and Pattern Recognition</subject><creationdate>2021-01</creationdate><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,780,885</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2101.02663$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2101.02663$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Vemparala, Manoj-Rohit</creatorcontrib><creatorcontrib>Fasfous, Nael</creatorcontrib><creatorcontrib>Frickenstein, Alexander</creatorcontrib><creatorcontrib>Moraly, Mhd Ali</creatorcontrib><creatorcontrib>Jamal, Aquib</creatorcontrib><creatorcontrib>Frickenstein, Lukas</creatorcontrib><creatorcontrib>Unger, Christian</creatorcontrib><creatorcontrib>Nagaraja, Naveen-Shankar</creatorcontrib><creatorcontrib>Stechele, Walter</creatorcontrib><title>L2PF -- Learning to Prune Faster</title><description>Various applications in the field of autonomous driving are based on convolutional neural networks (CNNs), especially for processing camera data. The optimization of such CNNs is a major challenge in continuous development. Newly learned features must be brought into vehicles as quickly as possible, and as such, it is not feasible to spend redundant GPU hours during compression. In this context, we present Learning to Prune Faster which details a multi-task, try-and-learn method, discretely learning redundant filters of the CNN and a continuous action of how long the layers have to be fine-tuned. This allows us to significantly speed up the convergence process of learning how to find an embedded-friendly filter-wise pruned CNN. For ResNet20, we have achieved a compression ratio of 3.84 x with minimal accuracy degradation. Compared to the state-of-the-art pruning method, we reduced the GPU hours by 1.71 x.</description><subject>Computer Science - Artificial Intelligence</subject><subject>Computer Science - Computer Vision and Pattern Recognition</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2021</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotzrEOgjAUQNEuDgb9ACf7A8XXlpYyGiJqQqKDO3mBV0OiaAoa_XsjOt3t5jC2kBAnzhhYYXi1z1hJkDEoa_WU8VIdCy4ELwlD13ZnPtz4MTw64gX2A4UZm3i89DT_N2KnYnPKd6I8bPf5uhRoUy2cUYasl1I3rnE2c76GWqeUKECfekegMsokSY_oG2kTlaInMInOQDtCHbHlbzsSq3torxje1ZdajVT9AY6qNjE</recordid><startdate>20210107</startdate><enddate>20210107</enddate><creator>Vemparala, Manoj-Rohit</creator><creator>Fasfous, Nael</creator><creator>Frickenstein, Alexander</creator><creator>Moraly, Mhd Ali</creator><creator>Jamal, Aquib</creator><creator>Frickenstein, Lukas</creator><creator>Unger, Christian</creator><creator>Nagaraja, Naveen-Shankar</creator><creator>Stechele, Walter</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20210107</creationdate><title>L2PF -- Learning to Prune Faster</title><author>Vemparala, Manoj-Rohit ; Fasfous, Nael ; Frickenstein, Alexander ; Moraly, Mhd Ali ; Jamal, Aquib ; Frickenstein, Lukas ; Unger, Christian ; Nagaraja, Naveen-Shankar ; Stechele, Walter</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a673-8525e6f113d8d8698fc0c37e420af7f8e029e91e1faafd16427afe05439038ea3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2021</creationdate><topic>Computer Science - Artificial Intelligence</topic><topic>Computer Science - Computer Vision and Pattern Recognition</topic><toplevel>online_resources</toplevel><creatorcontrib>Vemparala, Manoj-Rohit</creatorcontrib><creatorcontrib>Fasfous, Nael</creatorcontrib><creatorcontrib>Frickenstein, Alexander</creatorcontrib><creatorcontrib>Moraly, Mhd Ali</creatorcontrib><creatorcontrib>Jamal, Aquib</creatorcontrib><creatorcontrib>Frickenstein, Lukas</creatorcontrib><creatorcontrib>Unger, Christian</creatorcontrib><creatorcontrib>Nagaraja, Naveen-Shankar</creatorcontrib><creatorcontrib>Stechele, Walter</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Vemparala, Manoj-Rohit</au><au>Fasfous, Nael</au><au>Frickenstein, Alexander</au><au>Moraly, Mhd Ali</au><au>Jamal, Aquib</au><au>Frickenstein, Lukas</au><au>Unger, Christian</au><au>Nagaraja, Naveen-Shankar</au><au>Stechele, Walter</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>L2PF -- Learning to Prune Faster</atitle><date>2021-01-07</date><risdate>2021</risdate><abstract>Various applications in the field of autonomous driving are based on convolutional neural networks (CNNs), especially for processing camera data. The optimization of such CNNs is a major challenge in continuous development. Newly learned features must be brought into vehicles as quickly as possible, and as such, it is not feasible to spend redundant GPU hours during compression. In this context, we present Learning to Prune Faster which details a multi-task, try-and-learn method, discretely learning redundant filters of the CNN and a continuous action of how long the layers have to be fine-tuned. This allows us to significantly speed up the convergence process of learning how to find an embedded-friendly filter-wise pruned CNN. For ResNet20, we have achieved a compression ratio of 3.84 x with minimal accuracy degradation. Compared to the state-of-the-art pruning method, we reduced the GPU hours by 1.71 x.</abstract><doi>10.48550/arxiv.2101.02663</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2101.02663
ispartof
issn
language eng
recordid cdi_arxiv_primary_2101_02663
source arXiv.org
subjects Computer Science - Artificial Intelligence
Computer Science - Computer Vision and Pattern Recognition
title L2PF -- Learning to Prune Faster
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-12T23%3A43%3A50IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=L2PF%20--%20Learning%20to%20Prune%20Faster&rft.au=Vemparala,%20Manoj-Rohit&rft.date=2021-01-07&rft_id=info:doi/10.48550/arxiv.2101.02663&rft_dat=%3Carxiv_GOX%3E2101_02663%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true