Improving transfer learning accuracy by reusing Stacked Denoising Autoencoders

Transfer learning is a process that allows reusing a learning machine trained on a problem to solve a new problem. Transfer learning studies on shallow architectures show low performance as they are generally based on hand-crafted features obtained from experts. It is therefore interesting to study...

Full description

Saved in:
Bibliographic Details
Published in:Conference proceedings - IEEE International Conference on Systems, Man, and Cybernetics pp. 1380 - 1387
Main Authors: Kandaswamy, Chetak, Silva, Luis M., Alexandre, Luis A., Sousa, Ricardo, Santos, Jorge M., de Sa, Joaquim Marques
Format: Conference Proceeding
Language:English
Published: IEEE 01.10.2014
Subjects:
ISSN:1062-922X
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Abstract Transfer learning is a process that allows reusing a learning machine trained on a problem to solve a new problem. Transfer learning studies on shallow architectures show low performance as they are generally based on hand-crafted features obtained from experts. It is therefore interesting to study transference on deep architectures, known to directly extract the features from the input data. A Stacked Denoising Autoencoder (SDA) is a deep model able to represent the hierarchical features needed for solving classification problems. In this paper we study the performance of SDAs trained on one problem and reused to solve a different problem not only with different distribution but also with a different tasks. We propose two different approaches: 1) unsupervised feature transference, and 2) supervised feature transference using deep transfer learning. We show that SDAs using the unsupervised feature transference outperform randomly initialized machines on a new problem. We achieved 7% relative improvement on average error rate and 41% on average computation time to classify typed uppercase letters. In the case of supervised feature transference, we achieved 5.7% relative improvement in the average error rate, by reusing the first and second hidden layer, and 8.5% relative improvement for the average error rate and 54% speed up w.r.t the baseline by reusing all three hidden layers for the same data. We also explore transfer learning between geometrical shapes and canonical shapes, we achieved 7.4% relative improvement on average error rate in case of supervised feature transference approach.
AbstractList Transfer learning is a process that allows reusing a learning machine trained on a problem to solve a new problem. Transfer learning studies on shallow architectures show low performance as they are generally based on hand-crafted features obtained from experts. It is therefore interesting to study transference on deep architectures, known to directly extract the features from the input data. A Stacked Denoising Autoencoder (SDA) is a deep model able to represent the hierarchical features needed for solving classification problems. In this paper we study the performance of SDAs trained on one problem and reused to solve a different problem not only with different distribution but also with a different tasks. We propose two different approaches: 1) unsupervised feature transference, and 2) supervised feature transference using deep transfer learning. We show that SDAs using the unsupervised feature transference outperform randomly initialized machines on a new problem. We achieved 7% relative improvement on average error rate and 41% on average computation time to classify typed uppercase letters. In the case of supervised feature transference, we achieved 5.7% relative improvement in the average error rate, by reusing the first and second hidden layer, and 8.5% relative improvement for the average error rate and 54% speed up w.r.t the baseline by reusing all three hidden layers for the same data. We also explore transfer learning between geometrical shapes and canonical shapes, we achieved 7.4% relative improvement on average error rate in case of supervised feature transference approach.
Author Kandaswamy, Chetak
Silva, Luis M.
Sousa, Ricardo
de Sa, Joaquim Marques
Santos, Jorge M.
Alexandre, Luis A.
Author_xml – sequence: 1
  givenname: Chetak
  surname: Kandaswamy
  fullname: Kandaswamy, Chetak
  email: chetak.kand@gmail.com
  organization: Inst. de Eng. Biomed. (INEB), Portugal
– sequence: 2
  givenname: Luis M.
  surname: Silva
  fullname: Silva, Luis M.
  email: lmas@ua.pt
  organization: INEB, Porto, Portugal
– sequence: 3
  givenname: Luis A.
  surname: Alexandre
  fullname: Alexandre, Luis A.
  organization: Univ. da Beira Interior, Covilhã, Portugal
– sequence: 4
  givenname: Ricardo
  surname: Sousa
  fullname: Sousa, Ricardo
  organization: INEB, Univ. do Porto, Porto, Portugal
– sequence: 5
  givenname: Jorge M.
  surname: Santos
  fullname: Santos, Jorge M.
  organization: INEB, Porto, Portugal
– sequence: 6
  givenname: Joaquim Marques
  surname: de Sa
  fullname: de Sa, Joaquim Marques
  organization: INEB, Porto, Portugal
BookMark eNotz0FLwzAYxvEIE9zm7oKXfoHW903TJjmOqnMw9TAFbyNN3kp1S0fSCv32Mt3pgd_hgf-MTXznibEbhAwR9N32uco4oMhKLQWCvGALLRUKqXWuBMgJmyKUPNWcf1yxWYxfABwEqil7WR-Ooftp_WfSB-NjQyHZkwn-JMbaIRg7JvWYBBriyba9sd_kknvyXfsny6HvyNvOUYjX7LIx-0iL887Z--PDW_WUbl5X62q5SVuURZ82gjgq0tZpjiVZBFC2tooKJ3lJhTSaVIFcNxxdKcCZ2hneyEYi1rIw-Zzd_v-2RLQ7hvZgwrg75-e_I_hRZg
ContentType Conference Proceeding
DBID 6IE
6IH
CBEJK
RIE
RIO
DOI 10.1109/SMC.2014.6974107
DatabaseName IEEE Electronic Library (IEL) Conference Proceedings
IEEE Proceedings Order Plan (POP) 1998-present by volume
IEEE Xplore All Conference Proceedings
IEEE Electronic Library (IEL)
IEEE Proceedings Order Plans (POP) 1998-present
DatabaseTitleList
Database_xml – sequence: 1
  dbid: RIE
  name: IEEE Electronic Library (IEL)
  url: https://ieeexplore.ieee.org/
  sourceTypes: Publisher
DeliveryMethod fulltext_linktorsrc
Discipline Engineering
Sciences (General)
EISBN 9781479938407
1479938408
EndPage 1387
ExternalDocumentID 6974107
Genre orig-research
GroupedDBID 29F
6IE
6IF
6IH
6IK
6IL
6IM
6IN
AAJGR
AAWTH
ABLEC
ADZIZ
ALMA_UNASSIGNED_HOLDINGS
BEFXN
BFFAM
BGNUA
BKEBE
BPEOZ
CBEJK
CHZPO
IEGSK
IJVOP
IPLJI
M43
OCL
RIE
RIL
RIO
RNS
ID FETCH-LOGICAL-i175t-f4e218e9cd9216ec1008cbc8e5d726e57a9e85129f21d640dabda2f7f711b75a3
IEDL.DBID RIE
ISICitedReferencesCount 25
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000370963701084&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN 1062-922X
IngestDate Wed Aug 27 04:46:46 EDT 2025
IsPeerReviewed false
IsScholarly true
Language English
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-i175t-f4e218e9cd9216ec1008cbc8e5d726e57a9e85129f21d640dabda2f7f711b75a3
PageCount 8
ParticipantIDs ieee_primary_6974107
PublicationCentury 2000
PublicationDate 2014-Oct.
PublicationDateYYYYMMDD 2014-10-01
PublicationDate_xml – month: 10
  year: 2014
  text: 2014-Oct.
PublicationDecade 2010
PublicationTitle Conference proceedings - IEEE International Conference on Systems, Man, and Cybernetics
PublicationTitleAbbrev SMC
PublicationYear 2014
Publisher IEEE
Publisher_xml – name: IEEE
SSID ssj0020418
Score 2.0978892
Snippet Transfer learning is a process that allows reusing a learning machine trained on a problem to solve a new problem. Transfer learning studies on shallow...
SourceID ieee
SourceType Publisher
StartPage 1380
SubjectTerms Computer architecture
Deep Learning
Error analysis
Noise reduction
Shape
Training
Transfer Learning
Visualization
Yttrium
Title Improving transfer learning accuracy by reusing Stacked Denoising Autoencoders
URI https://ieeexplore.ieee.org/document/6974107
WOSCitedRecordID wos000370963701084&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
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV3LSsQwFA3j4EI3Og_xTRYuFOxMk7ZJs5TRwYUOAz6Y3ZDHjcymlU4rzN_btLUquHEXAoGQS7gnueeci9CFLzTRIgg9GwfWC2MKniSKeVGkeMCYtn5lX_z6wGezeLEQ8w66brUwAFCRz2DkhlUt36S6cF9lY1aCX-Kk41ucs1qr1T6u_JDUsjdGPUHp4qsk6Yvx0-PEcbjCUbP-VyOVKo9M9_63g300_Bbk4XmbanqoA0kf7f7wEuyjXnNL1_iysZK-GqBZ-2eA8wqhQoabPhFvWGpdZFJvsNrgDBwB_g2X2LO81gbfQpKuqpmbIk-d2aUjPA_Ry_TueXLvNR0UvFUJC3LPhlCmcBDaCEoYaOfko5WOITKcMoi4FBC7lG8pMSz0jVRGUsstJ0TxSAYHqJukCRwirCwxpnzR-kLaUAqtrFJEMhsaoamv9REauONavtcmGcvmpI7_nj5BOy4iNSvuFHXzrIAztK0_8tU6O68i-wkkyKUT
linkProvider IEEE
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1NSwMxEA2lCupF-yF-m4MHBbfdpNns5ihqqdiWglV6K_mYlF66st0V-u_dbNeq4MVbyCGEDOFNMu-9QejKF5po0WGejTrWYxEFTxLFvSBQYYdzbf3CvvitHw6H0WQiRhV0u9HCAEBBPoOWGxa1fBPrzH2VtXme_BInHd9ynbNKtdbmeeUzsha-ceoJSidfRUlftF8G947FxVrlCr9aqRRI0t3_3x4OUPNbkodHG7CpoQos6mjvh5tgHdXKe7rE16WZ9E0DDTe_BjgtclRIcNkpYoal1lki9QqrFU7AUeBnOM8-84tt8AMs4nkxc5elsbO7dJTnJnrtPo7ve17ZQ8Gb54lB6lkGOYiD0EZQwkE7Lx-tdASBCSmHIJQCIgf6lhLDmW-kMpLa0IaEqDCQnUNUXcQLOEJYWWJM_qb1hbRMCq2sUkRyy4zQ1Nf6GDXccU3f1zYZ0_KkTv6evkQ7vfGgP-0_DZ9P0a6Lzpojd4aqaZLBOdrWH-l8mVwUUf4E6-WoXA
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%3Abook&rft.genre=proceeding&rft.title=Conference+proceedings+-+IEEE+International+Conference+on+Systems%2C+Man%2C+and+Cybernetics&rft.atitle=Improving+transfer+learning+accuracy+by+reusing+Stacked+Denoising+Autoencoders&rft.au=Kandaswamy%2C+Chetak&rft.au=Silva%2C+Luis+M.&rft.au=Alexandre%2C+Luis+A.&rft.au=Sousa%2C+Ricardo&rft.date=2014-10-01&rft.pub=IEEE&rft.issn=1062-922X&rft.spage=1380&rft.epage=1387&rft_id=info:doi/10.1109%2FSMC.2014.6974107&rft.externalDocID=6974107
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=1062-922X&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=1062-922X&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=1062-922X&client=summon