DeepCABAC: A Universal Compression Algorithm for Deep Neural Networks

In the past decade deep neural networks (DNNs) have shown state-of-the-art performance on a wide range of complex machine learning tasks. Many of these results have been achieved while growing the size of DNNs, creating a demand for efficient compression and transmission of them. In this work we pre...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:IEEE journal of selected topics in signal processing Ročník 14; číslo 4; s. 700 - 714
Hlavní autori: Wiedemann, Simon, Kirchhoffer, Heiner, Matlage, Stefan, Haase, Paul, Marban, Arturo, Marinc, Talmaj, Neumann, David, Nguyen, Tung, Schwarz, Heiko, Wiegand, Thomas, Marpe, Detlev, Samek, Wojciech
Médium: Journal Article
Jazyk:English
Vydavateľské údaje: New York IEEE 01.05.2020
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Predmet:
ISSN:1932-4553, 1941-0484
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Abstract In the past decade deep neural networks (DNNs) have shown state-of-the-art performance on a wide range of complex machine learning tasks. Many of these results have been achieved while growing the size of DNNs, creating a demand for efficient compression and transmission of them. In this work we present DeepCABAC, a universal compression algorithm for DNNs that is based on applying Context-based Adaptive Binary Arithmetic Coder (CABAC) to the DNN parameters. CABAC was originally designed for the H.264/AVC video coding standard and became the state-of-the-art for the lossless compression part of video compression. DeepCABAC applies a novel quantization scheme that minimizes a rate-distortion function while simultaneously taking the impact of quantization to the DNN performance into account. Experimental results show that DeepCABAC consistently attains higher compression rates than previously proposed coding techniques for DNN compression. For instance, it is able to compress the VGG16 ImageNet model by x63.6 with no loss of accuracy, thus being able to represent the entire network with merely 9 MB. The source code for encoding and decoding can be found at https://github.com/fraunhoferhhi/DeepCABAC .
AbstractList In the past decade deep neural networks (DNNs) have shown state-of-the-art performance on a wide range of complex machine learning tasks. Many of these results have been achieved while growing the size of DNNs, creating a demand for efficient compression and transmission of them. In this work we present DeepCABAC, a universal compression algorithm for DNNs that is based on applying Context-based Adaptive Binary Arithmetic Coder (CABAC) to the DNN parameters. CABAC was originally designed for the H.264/AVC video coding standard and became the state-of-the-art for the lossless compression part of video compression. DeepCABAC applies a novel quantization scheme that minimizes a rate-distortion function while simultaneously taking the impact of quantization to the DNN performance into account. Experimental results show that DeepCABAC consistently attains higher compression rates than previously proposed coding techniques for DNN compression. For instance, it is able to compress the VGG16 ImageNet model by x63.6 with no loss of accuracy, thus being able to represent the entire network with merely 9 MB. The source code for encoding and decoding can be found at https://github.com/fraunhoferhhi/DeepCABAC .
In the past decade deep neural networks (DNNs) have shown state-of-the-art performance on a wide range of complex machine learning tasks. Many of these results have been achieved while growing the size of DNNs, creating a demand for efficient compression and transmission of them. In this work we present DeepCABAC, a universal compression algorithm for DNNs that is based on applying Context-based Adaptive Binary Arithmetic Coder (CABAC) to the DNN parameters. CABAC was originally designed for the H.264/AVC video coding standard and became the state-of-the-art for the lossless compression part of video compression. DeepCABAC applies a novel quantization scheme that minimizes a rate-distortion function while simultaneously taking the impact of quantization to the DNN performance into account. Experimental results show that DeepCABAC consistently attains higher compression rates than previously proposed coding techniques for DNN compression. For instance, it is able to compress the VGG16 ImageNet model by x63.6 with no loss of accuracy, thus being able to represent the entire network with merely 9 MB. The source code for encoding and decoding can be found at https://github.com/fraunhoferhhi/DeepCABAC .
Author Marban, Arturo
Neumann, David
Nguyen, Tung
Haase, Paul
Wiedemann, Simon
Matlage, Stefan
Marpe, Detlev
Samek, Wojciech
Schwarz, Heiko
Kirchhoffer, Heiner
Marinc, Talmaj
Wiegand, Thomas
Author_xml – sequence: 1
  givenname: Simon
  orcidid: 0000-0001-5144-3758
  surname: Wiedemann
  fullname: Wiedemann, Simon
  email: simon.wiedemann@hhi.fraunhofer.de
  organization: Fraunhofer Heinrich Hertz Institute, Berlin, Germany
– sequence: 2
  givenname: Heiner
  surname: Kirchhoffer
  fullname: Kirchhoffer, Heiner
  email: heiner.kirchhoffer@hhi.fraunhofer.de
  organization: Fraunhofer Heinrich Hertz Institute, Berlin, Germany
– sequence: 3
  givenname: Stefan
  surname: Matlage
  fullname: Matlage, Stefan
  email: stefan.matlage@hhi.fraunhofer.de
  organization: Fraunhofer Heinrich Hertz Institute, Berlin, Germany
– sequence: 4
  givenname: Paul
  orcidid: 0000-0002-0273-4564
  surname: Haase
  fullname: Haase, Paul
  email: paul.haase@hhi.fraunhofer.de
  organization: Fraunhofer Heinrich Hertz Institute, Berlin, Germany
– sequence: 5
  givenname: Arturo
  surname: Marban
  fullname: Marban, Arturo
  email: arturo.marban@hhi-extern.fraunhofer.de
  organization: Fraunhofer Heinrich Hertz Institute, Berlin, Germany
– sequence: 6
  givenname: Talmaj
  surname: Marinc
  fullname: Marinc, Talmaj
  email: talmaj.marinc@hhi.fraunhofer.de
  organization: Fraunhofer Heinrich Hertz Institute, Berlin, Germany
– sequence: 7
  givenname: David
  surname: Neumann
  fullname: Neumann, David
  email: david.neumann@hhi.fraunhofer.de
  organization: Fraunhofer Heinrich Hertz Institute, Berlin, Germany
– sequence: 8
  givenname: Tung
  orcidid: 0000-0002-7429-4932
  surname: Nguyen
  fullname: Nguyen, Tung
  email: tung.nguyen@hhi.fraunhofer.de
  organization: Fraunhofer Heinrich Hertz Institute, Berlin, Germany
– sequence: 9
  givenname: Heiko
  orcidid: 0000-0002-7136-0041
  surname: Schwarz
  fullname: Schwarz, Heiko
  email: heiko.schwarz@hhi.fraunhofer.de
  organization: Fraunhofer Heinrich Hertz Institute, Berlin, Germany
– sequence: 10
  givenname: Thomas
  surname: Wiegand
  fullname: Wiegand, Thomas
  email: Thomas.Wiegand@hhi.fraunhofer.de
  organization: Fraunhofer Heinrich Hertz Institute, Berlin, Germany
– sequence: 11
  givenname: Detlev
  orcidid: 0000-0002-5391-3247
  surname: Marpe
  fullname: Marpe, Detlev
  email: detlev.marpe@hhi.fraunhofer.de
  organization: Fraunhofer Heinrich Hertz Institute, Berlin, Germany
– sequence: 12
  givenname: Wojciech
  orcidid: 0000-0002-6283-3265
  surname: Samek
  fullname: Samek, Wojciech
  email: wojciech.samek@hhi.fraunhofer.de
  organization: Fraunhofer Heinrich Hertz Institute, Berlin, Germany
BookMark eNp9kMtOwzAQRS0EEm3hB2ATiXWKn7HNLoTyUlWQ2q6tNB1DShoXOwXx9yS0YsGC1czinjua00eHtasBoTOCh4Rgffk4nU2fhxRTPKQ60ULwA9QjmpMYc8UPu53RmAvBjlE_hBXGQiaE99DoBmCTpddpdhWl0bwuP8CHvIoyt954CKF0dZRWL86Xzes6ss5HHRBNYOvb1ASaT-ffwgk6snkV4HQ_B2h-O5pl9_H46e4hS8dxwbFoYrbkAnLCrVSKJ6SwWgqllgnghV6yhFrF-GKRU13wRIIWiVIWJMuFoJJbJdkAXex6N969byE0ZuW2vm5PGsoZJ1QS0aXULlV4F4IHa4qyyZv2lcbnZWUINp008yPNdNLMXlqL0j_oxpfr3H_9D53voBIAfgGlJaaas2_d1Xhv
CODEN IJSTGY
CitedBy_id crossref_primary_10_1109_JSTSP_2023_3312914
crossref_primary_10_1109_TCSVT_2023_3302858
crossref_primary_10_1145_3634686
crossref_primary_10_1155_2022_4886586
crossref_primary_10_3390_electronics12194042
crossref_primary_10_1016_j_cosrev_2023_100568
crossref_primary_10_1109_TCSVT_2024_3467124
crossref_primary_10_1371_journal_pone_0303462
crossref_primary_10_1007_s00500_023_09109_5
crossref_primary_10_3390_app11083725
crossref_primary_10_1145_3588436
crossref_primary_10_1016_j_asoc_2024_111901
crossref_primary_10_1016_j_ins_2024_120367
crossref_primary_10_1109_ACCESS_2025_3582850
crossref_primary_10_1088_1361_6579_acb24d
crossref_primary_10_1109_JIOT_2023_3340858
crossref_primary_10_1145_3607139
crossref_primary_10_1016_j_procs_2022_12_239
crossref_primary_10_1016_j_neucom_2025_130359
crossref_primary_10_1109_JSTSP_2022_3142678
crossref_primary_10_1088_1742_6596_2216_1_012078
crossref_primary_10_1109_TSC_2023_3336980
crossref_primary_10_3390_make3020020
crossref_primary_10_1109_JIOT_2022_3197317
crossref_primary_10_1109_TNNLS_2020_3015958
crossref_primary_10_23919_JSEE_2023_000159
crossref_primary_10_1038_s41467_020_19203_z
crossref_primary_10_1109_LSP_2021_3052097
crossref_primary_10_3390_drones8120707
crossref_primary_10_1109_TII_2022_3192882
crossref_primary_10_3390_s22229003
crossref_primary_10_1186_s40537_021_00444_8
crossref_primary_10_1016_j_displa_2024_102880
crossref_primary_10_1109_TNSE_2021_3081748
crossref_primary_10_1016_j_comcom_2024_07_012
crossref_primary_10_1109_TMC_2022_3190260
crossref_primary_10_1109_ACCESS_2024_3467375
crossref_primary_10_1109_TCSVT_2022_3229079
crossref_primary_10_1109_LSP_2025_3578291
crossref_primary_10_1109_TMM_2024_3357198
crossref_primary_10_1109_TCSVT_2021_3095970
crossref_primary_10_1109_TNNLS_2021_3063265
crossref_primary_10_1109_TWC_2024_3388329
Cites_doi 10.1109/TC.2019.2914438
10.1038/s41928-018-0059-3
10.1109/MSP.2017.2765695
10.1561/2000000010
10.1002/j.1538-7305.1948.tb01338.x
10.1109/ICIP.2003.1246667
10.1109/TNNLS.2019.2910073
10.1109/IJCNN.2019.8852172
10.1007/978-3-319-06895-4
10.1109/JRPROC.1952.273898
10.1145/3092831
10.1038/s41467-019-08987-4
10.7551/mitpress/4643.001.0001
10.1109/CVPR.2019.01099
10.1038/nature14539
10.1109/ICCV.2019.00141
10.1016/0020-0190(78)90024-8
10.1109/TCSVT.2003.815173
10.1007/978-3-319-46493-0_32
10.1109/IJCNN.2017.7966159
10.1109/ICNN.1993.298572
10.1109/CVPR.2019.01166
10.1109/TNNLS.2019.2944481
10.1109/IJCNN.2019.8852119
ContentType Journal Article
Copyright Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2020
Copyright_xml – notice: Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2020
DBID 97E
ESBDL
RIA
RIE
AAYXX
CITATION
7SP
8FD
H8D
L7M
DOI 10.1109/JSTSP.2020.2969554
DatabaseName IEEE Xplore (IEEE)
IEEE Xplore Open Access Journals
IEEE All-Society Periodicals Package (ASPP) 1998–Present
IEEE Electronic Library (IEL)
CrossRef
Electronics & Communications Abstracts
Technology Research Database
Aerospace Database
Advanced Technologies Database with Aerospace
DatabaseTitle CrossRef
Aerospace Database
Technology Research Database
Advanced Technologies Database with Aerospace
Electronics & Communications Abstracts
DatabaseTitleList Aerospace Database

Database_xml – sequence: 1
  dbid: RIE
  name: IEEE Electronic Library (IEL)
  url: https://ieeexplore.ieee.org/
  sourceTypes: Publisher
DeliveryMethod fulltext_linktorsrc
Discipline Engineering
EISSN 1941-0484
EndPage 714
ExternalDocumentID 10_1109_JSTSP_2020_2969554
8970294
Genre orig-research
GrantInformation_xml – fundername: Berlin Center for Machine Learning
  grantid: 01IS18037I
– fundername: German Ministry for Education through the Berlin Big Data Center
  grantid: 01IS14013A
GroupedDBID -~X
0R~
29I
4.4
5GY
5VS
6IK
97E
AAJGR
AARMG
AASAJ
AAWTH
ABAZT
ABQJQ
ABVLG
ACIWK
AENEX
AETIX
AGQYO
AGSQL
AHBIQ
AKJIK
AKQYR
ALMA_UNASSIGNED_HOLDINGS
ATWAV
BEFXN
BFFAM
BGNUA
BKEBE
BPEOZ
CS3
DU5
EBS
EJD
ESBDL
F5P
HZ~
IFIPE
IPLJI
JAVBF
LAI
M43
O9-
OCL
RIA
RIE
RNS
AAYXX
CITATION
7SP
8FD
H8D
L7M
ID FETCH-LOGICAL-c405t-3d45ea14f788461cf97588d6e0b9d362f834bba29c467e95688fe73a55274f873
IEDL.DBID RIE
ISICitedReferencesCount 75
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000565858900009&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN 1932-4553
IngestDate Mon Jun 30 10:19:44 EDT 2025
Sat Nov 29 04:10:33 EST 2025
Tue Nov 18 22:42:23 EST 2025
Wed Aug 27 02:32:37 EDT 2025
IsDoiOpenAccess true
IsOpenAccess true
IsPeerReviewed true
IsScholarly true
Issue 4
Language English
License https://creativecommons.org/licenses/by/4.0/legalcode
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-c405t-3d45ea14f788461cf97588d6e0b9d362f834bba29c467e95688fe73a55274f873
Notes ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ORCID 0000-0002-7136-0041
0000-0001-5144-3758
0000-0002-7429-4932
0000-0002-0273-4564
0000-0002-6283-3265
0000-0002-5391-3247
OpenAccessLink https://ieeexplore.ieee.org/document/8970294
PQID 2434127157
PQPubID 75721
PageCount 15
ParticipantIDs crossref_citationtrail_10_1109_JSTSP_2020_2969554
ieee_primary_8970294
proquest_journals_2434127157
crossref_primary_10_1109_JSTSP_2020_2969554
PublicationCentury 2000
PublicationDate 2020-05-01
PublicationDateYYYYMMDD 2020-05-01
PublicationDate_xml – month: 05
  year: 2020
  text: 2020-05-01
  day: 01
PublicationDecade 2020
PublicationPlace New York
PublicationPlace_xml – name: New York
PublicationTitle IEEE journal of selected topics in signal processing
PublicationTitleAbbrev JSTSP
PublicationYear 2020
Publisher IEEE
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Publisher_xml – name: IEEE
– name: The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
References ref13
guo (ref29) 2016
ref15
ref14
hinton (ref44) 2015
zhou (ref39) 2017
ref11
ref10
howard (ref4) 2017
choi (ref36) 2018
ref17
ref19
ref18
ref51
cheng (ref9) 2017
louizos (ref33) 0
(ref26) 2019
(ref46) 2019
han (ref32) 2015
ref48
ref47
(ref25) 2019
wiegand (ref16) 2011; 4
ref49
han (ref28) 0
ref8
ref7
chetlur (ref20) 2014
mcmahan (ref6) 2016
ref5
courbariaux (ref37) 2016
molchanov (ref22) 0
sattler (ref50) 2019
ref34
xiaowei (ref2) 2018; 1
achille (ref23) 2017
ref38
horowitz (ref3) 0
devlin (ref42) 2018
lan (ref43) 2019
choi (ref35) 2016
ref24
redmon (ref41) 2018
lecun (ref1) 2015; 521
zhu (ref31) 0
ref21
ref27
rastegari (ref30) 2016
(ref12) 2019
simonyan (ref40) 2014
tan (ref45) 2019; abs 1907 9595
References_xml – year: 2017
  ident: ref39
  article-title: Incremental network quantization: Towards lossless CNNs with low-precision weights
– year: 2015
  ident: ref32
  article-title: Deep compression: Compressing deep neural network with pruning, trained quantization and huffman coding
– year: 2018
  ident: ref41
  article-title: Yolov3: An incremental improvement
– year: 2019
  ident: ref46
  article-title: Working draft 2 of compression of neural networks for multimedia content description and analysis (n18784)
– ident: ref51
  doi: 10.1109/TC.2019.2914438
– year: 2017
  ident: ref9
  article-title: A survey of model compression and acceleration for deep neural networks
– volume: 1
  start-page: 216
  year: 2018
  ident: ref2
  article-title: Scaling for edge inference of deep neural networks
  publication-title: Nature Electron
  doi: 10.1038/s41928-018-0059-3
– year: 2014
  ident: ref20
  article-title: cuDNN: Efficient primitives for deep learning
– ident: ref10
  doi: 10.1109/MSP.2017.2765695
– volume: 4
  start-page: 1
  year: 2011
  ident: ref16
  article-title: Source coding: Part 1 of fundamentals of source and video coding
  publication-title: Found Trends Signal Process
  doi: 10.1561/2000000010
– ident: ref15
  doi: 10.1002/j.1538-7305.1948.tb01338.x
– ident: ref18
  doi: 10.1109/ICIP.2003.1246667
– ident: ref11
  doi: 10.1109/TNNLS.2019.2910073
– ident: ref7
  doi: 10.1109/IJCNN.2019.8852172
– year: 0
  ident: ref31
  article-title: Trained ternary quantization
  publication-title: Proc Int Conf Learn Representations
– ident: ref13
  doi: 10.1007/978-3-319-06895-4
– ident: ref17
  doi: 10.1109/JRPROC.1952.273898
– year: 2016
  ident: ref6
  article-title: Federated learning of deep networks using model averaging
– ident: ref5
  doi: 10.1145/3092831
– ident: ref49
  doi: 10.1038/s41467-019-08987-4
– year: 2018
  ident: ref42
  article-title: BERT: Pre-training of deep bidirectional transformers for language understanding
– ident: ref19
  doi: 10.7551/mitpress/4643.001.0001
– start-page: 10
  year: 0
  ident: ref3
  article-title: 1.1 computing's energy problem (and what we can do about it)
  publication-title: Proc IEEE Int Solid-State Circuits Conf Dig Tech Papers
– year: 2019
  ident: ref26
  article-title: Test model 2 of compression of neural networks for multimedia content description and analysis (n18785)
– year: 2019
  ident: ref12
  article-title: Updated call for proposals on neural network compression (n18129)
– start-page: 2498
  year: 0
  ident: ref22
  article-title: Variational dropout sparsifies deep neural networks
  publication-title: Proc Int Conf Mach Learn
– ident: ref47
  doi: 10.1109/CVPR.2019.01099
– year: 2019
  ident: ref50
  article-title: Clustered federated learning: Model-agnostic distributed multi-task optimization under privacy constraints
– volume: 521
  start-page: 436
  year: 2015
  ident: ref1
  article-title: Deep learning
  publication-title: Nature
  doi: 10.1038/nature14539
– start-page: 1135
  year: 0
  ident: ref28
  article-title: Learning both weights and connections for efficient neural networks
  publication-title: Proc Adv Neural Inf Process Syst
– volume: abs 1907 9595
  year: 2019
  ident: ref45
  article-title: Mixconv: Mixed depthwise convolutional kernels
  publication-title: CoRR
– year: 2016
  ident: ref37
  article-title: Binarynet: Training deep neural networks with weights and activations constrained to +1 or ?1
– year: 2017
  ident: ref4
  article-title: Mobilenets: Efficient convolutional neural networks for mobile vision applications
– ident: ref24
  doi: 10.1109/ICCV.2019.00141
– start-page: 3288
  year: 0
  ident: ref33
  article-title: Bayesian Compression for Deep Learning
  publication-title: Proc Adv Neural Inf Process Syst
– ident: ref21
  doi: 10.1016/0020-0190(78)90024-8
– year: 2014
  ident: ref40
  article-title: Very deep convolutional networks for large-scale image recognition
– ident: ref14
  doi: 10.1109/TCSVT.2003.815173
– year: 2016
  ident: ref30
  article-title: XNOR-Net: Imagenet classification using binary convolutional neural networks
  doi: 10.1007/978-3-319-46493-0_32
– year: 2019
  ident: ref43
  article-title: Albert: A lite BERT for self-supervised learning of language representations
– ident: ref38
  doi: 10.1109/IJCNN.2017.7966159
– year: 2016
  ident: ref35
  article-title: Towards the limit of network quantization
– ident: ref27
  doi: 10.1109/ICNN.1993.298572
– ident: ref48
  doi: 10.1109/CVPR.2019.01166
– ident: ref8
  doi: 10.1109/TNNLS.2019.2944481
– year: 2017
  ident: ref23
  article-title: Critical learning periods in deep neural networks
– year: 2018
  ident: ref36
  article-title: Universal deep neural network compression
– year: 2015
  ident: ref44
  article-title: Distilling the knowledge in a neural network
– year: 2016
  ident: ref29
  article-title: Dynamic network surgery for efficient dnns
– year: 2019
  ident: ref25
  article-title: Description of core experiments on compression of neural networks for multimedia content description and analysis (n18782)
– ident: ref34
  doi: 10.1109/IJCNN.2019.8852119
SSID ssj0057614
Score 2.5858557
Snippet In the past decade deep neural networks (DNNs) have shown state-of-the-art performance on a wide range of complex machine learning tasks. Many of these results...
SourceID proquest
crossref
ieee
SourceType Aggregation Database
Enrichment Source
Index Database
Publisher
StartPage 700
SubjectTerms Algorithms
arithmetic coding
Artificial neural networks
Coding
Coding standards
Cognitive tasks
Compression algorithms
Decoding
Deep learning
efficient representation
Machine learning
Measurement
neural network compression
Neural networks
Quantization (signal)
rate-distortion quantization
Source code
Source coding
Task complexity
Video compression
Title DeepCABAC: A Universal Compression Algorithm for Deep Neural Networks
URI https://ieeexplore.ieee.org/document/8970294
https://www.proquest.com/docview/2434127157
Volume 14
WOSCitedRecordID wos000565858900009&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
journalDatabaseRights – providerCode: PRVIEE
  databaseName: IEEE Electronic Library (IEL)
  customDbUrl:
  eissn: 1941-0484
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0057614
  issn: 1932-4553
  databaseCode: RIE
  dateStart: 20070101
  isFulltext: true
  titleUrlDefault: https://ieeexplore.ieee.org/
  providerName: IEEE
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV3NT8IwFH9B4kEPfqERRdODNx2sXbe23iZCPBhCoibcltG1SoKDwPDvt-0G0WhMvO3w3tL82r33e-v7ALjSzFcSZ4a5hSn3aEq0x6nIPMs2MGEiVKmbWvLIBgM-GolhDW42tTBKKZd8ptr20d3lZzO5sr_KOlwwnwi6BVuMRWWt1trqGtqMqxtk4tEwDNYFMr7omCP-NDShIPHbREQiDOk3J-Smqvwwxc6_9Pf_t7ID2Kt4JIrLjT-EmsqPYPdLd8EG9O6Vmnfju7h7i2JUZWAYHWsCyuzXHMXT19liUry9I0NekVVAtl2HkRqU-eHLY3jp9567D141NcGThnwVXpBRAzCm2gS3NMJSCxMS8CxS_lhkxl1pHtDxOCVCGhupbLUg14oFqW3FRjVnwQnU81muTgFFWkfKMC5MXds3yqXQUSppJnSqNMmagNcwJrJqKW4nW0wTF1r4InHQJxb6pIK-CdcbnXnZUONP6YYFeyNZ4dyE1nq3kuqbWyaEmkUShkN29rvWOezYd5fpii2oF4uVuoBt-VFMlotLd5w-AWP5xTY
linkProvider IEEE
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV3NT8IwFH9BNFEPfhtR1B286WDruq31NhGCERcSMeG2jK5VEhwEhn-_bbcRjcbE2w7vZc2v3Xu_t74PgCvhW5zZiWRubkxMHCNhEkwTU7ENG_nU5bGeWtLzw5AMh7RfgZtVLQznXCef8YZ61Hf5yZQt1a-yJqG-hSheg3U1OcvNq7VKuyuJs13cISMTu65TlshYtCkP-XNfBoPIaiDqUdfF39yQnqvywxhrD9PZ_d_a9mCnYJJGkG_9PlR4egDbX_oLHkL7nvNZK7gLWrdGYBQ5GFJHGYE8_zU1gsnrdD7O3t4NSV8NpWCohh1SKswzxBdH8NJpD1pds5ibYDJJvzLTSbCE2MZChrfYs5mgMiggicetEU2kwxLEwaNRjCiTVpKrekEiuO_EqhkbFsR3jqGaTlN-AoYnhMcl57KxbvyGCaPCixlOqIi5QEkN7BLGiBVNxdVsi0mkgwuLRhr6SEEfFdDX4HqlM8tbavwpfajAXkkWONegXu5WVHx1iwhhuUjk265_-rvWJWx2B0-9qPcQPp7BlnpPnrxYh2o2X_Jz2GAf2Xgxv9BH6xPdD8iB
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=DeepCABAC%3A+A+Universal+Compression+Algorithm+for+Deep+Neural+Networks&rft.jtitle=IEEE+journal+of+selected+topics+in+signal+processing&rft.au=Wiedemann%2C+Simon&rft.au=Kirchhoffer%2C+Heiner&rft.au=Matlage%2C+Stefan&rft.au=Haase%2C+Paul&rft.date=2020-05-01&rft.issn=1932-4553&rft.eissn=1941-0484&rft.volume=14&rft.issue=4&rft.spage=700&rft.epage=714&rft_id=info:doi/10.1109%2FJSTSP.2020.2969554&rft.externalDBID=n%2Fa&rft.externalDocID=10_1109_JSTSP_2020_2969554
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=1932-4553&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=1932-4553&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=1932-4553&client=summon