A robust variational autoencoder using beta divergence
The presence of outliers can severely degrade learned representations and performance of deep learning methods and hence disproportionately affect the training process, leading to incorrect conclusions about the data. For example, anomaly detection using deep generative models is typically only poss...
Uloženo v:
| Vydáno v: | Knowledge-based systems Ročník 238; s. 107886 |
|---|---|
| Hlavní autoři: | , , , , |
| Médium: | Journal Article |
| Jazyk: | angličtina |
| Vydáno: |
Netherlands
Elsevier B.V
28.02.2022
Elsevier Science Ltd |
| Témata: | |
| ISSN: | 0950-7051, 1872-7409 |
| On-line přístup: | Získat plný text |
| Tagy: |
Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
|
| Abstract | The presence of outliers can severely degrade learned representations and performance of deep learning methods and hence disproportionately affect the training process, leading to incorrect conclusions about the data. For example, anomaly detection using deep generative models is typically only possible when similar anomalies (or outliers) are not present in the training data. Here we focus on variational autoencoders (VAEs). While the VAE is a popular framework for anomaly detection tasks, we observe that the VAE is unable to detect outliers when the training data contains anomalies that have the same distribution as those in test data. In this paper we focus on robustness to outliers in training data in VAE settings using concepts from robust statistics. We propose a variational lower bound that leads to a robust VAE model that has the same computational complexity as the standard VAE and contains a single automatically-adjusted tuning parameter to control the degree of robustness. We present mathematical formulations for robust variational autoencoders (RVAEs) for Bernoulli, Gaussian and categorical variables. The RVAE model is based on beta-divergence rather than the standard Kullback–Leibler (KL) divergence. We demonstrate the performance of our proposed β-divergence-based autoencoder for a variety of image and categorical datasets showing improved robustness to outliers both qualitatively and quantitatively. We also illustrate the use of our robust VAE for detection of lesions in brain images, formulated as an anomaly detection task. Finally, we suggest a method to tune the hyperparameter of RVAE which makes our model completely unsupervised. |
|---|---|
| AbstractList | The presence of outliers can severely degrade learned representations and performance of deep learning methods and hence disproportionately affect the training process, leading to incorrect conclusions about the data. For example, anomaly detection using deep generative models is typically only possible when similar anomalies (or outliers) are not present in the training data. Here we focus on variational autoencoders (VAEs). While the VAE is a popular framework for anomaly detection tasks, we observe that the VAE is unable to detect outliers when the training data contains anomalies that have the same distribution as those in test data. In this paper we focus on robustness to outliers in training data in VAE settings using concepts from robust statistics. We propose a variational lower bound that leads to a robust VAE model that has the same computational complexity as the standard VAE and contains a single automatically-adjusted tuning parameter to control the degree of robustness. We present mathematical formulations for robust variational autoencoders (RVAEs) for Bernoulli, Gaussian and categorical variables. The RVAE model is based on beta-divergence rather than the standard Kullback-Leibler (KL) divergence. We demonstrate the performance of our proposed
-divergence-based autoencoder for a variety of image and categorical datasets showing improved robustness to outliers both qualitatively and quantitatively. We also illustrate the use of our robust VAE for detection of lesions in brain images, formulated as an anomaly detection task. Finally, we suggest a method to tune the hyperparameter of RVAE which makes our model completely unsupervised. The presence of outliers can severely degrade learned representations and performance of deep learning methods and hence disproportionately affect the training process, leading to incorrect conclusions about the data. For example, anomaly detection using deep generative models is typically only possible when similar anomalies (or outliers) are not present in the training data. Here we focus on variational autoencoders (VAEs). While the VAE is a popular framework for anomaly detection tasks, we observe that the VAE is unable to detect outliers when the training data contains anomalies that have the same distribution as those in test data. In this paper we focus on robustness to outliers in training data in VAE settings using concepts from robust statistics. We propose a variational lower bound that leads to a robust VAE model that has the same computational complexity as the standard VAE and contains a single automatically-adjusted tuning parameter to control the degree of robustness. We present mathematical formulations for robust variational autoencoders (RVAEs) for Bernoulli, Gaussian and categorical variables. The RVAE model is based on beta-divergence rather than the standard Kullback–Leibler (KL) divergence. We demonstrate the performance of our proposed β-divergence-based autoencoder for a variety of image and categorical datasets showing improved robustness to outliers both qualitatively and quantitatively. We also illustrate the use of our robust VAE for detection of lesions in brain images, formulated as an anomaly detection task. Finally, we suggest a method to tune the hyperparameter of RVAE which makes our model completely unsupervised. The presence of outliers can severely degrade learned representations and performance of deep learning methods and hence disproportionately affect the training process, leading to incorrect conclusions about the data. For example, anomaly detection using deep generative models is typically only possible when similar anomalies (or outliers) are not present in the training data. Here we focus on variational autoencoders (VAEs). While the VAE is a popular framework for anomaly detection tasks, we observe that the VAE is unable to detect outliers when the training data contains anomalies that have the same distribution as those in test data. In this paper we focus on robustness to outliers in training data in VAE settings using concepts from robust statistics. We propose a variational lower bound that leads to a robust VAE model that has the same computational complexity as the standard VAE and contains a single automatically-adjusted tuning parameter to control the degree of robustness. We present mathematical formulations for robust variational autoencoders (RVAEs) for Bernoulli, Gaussian and categorical variables. The RVAE model is based on beta-divergence rather than the standard Kullback-Leibler (KL) divergence. We demonstrate the performance of our proposed β-divergence-based autoencoder for a variety of image and categorical datasets showing improved robustness to outliers both qualitatively and quantitatively. We also illustrate the use of our robust VAE for detection of lesions in brain images, formulated as an anomaly detection task. Finally, we suggest a method to tune the hyperparameter of RVAE which makes our model completely unsupervised.The presence of outliers can severely degrade learned representations and performance of deep learning methods and hence disproportionately affect the training process, leading to incorrect conclusions about the data. For example, anomaly detection using deep generative models is typically only possible when similar anomalies (or outliers) are not present in the training data. Here we focus on variational autoencoders (VAEs). While the VAE is a popular framework for anomaly detection tasks, we observe that the VAE is unable to detect outliers when the training data contains anomalies that have the same distribution as those in test data. In this paper we focus on robustness to outliers in training data in VAE settings using concepts from robust statistics. We propose a variational lower bound that leads to a robust VAE model that has the same computational complexity as the standard VAE and contains a single automatically-adjusted tuning parameter to control the degree of robustness. We present mathematical formulations for robust variational autoencoders (RVAEs) for Bernoulli, Gaussian and categorical variables. The RVAE model is based on beta-divergence rather than the standard Kullback-Leibler (KL) divergence. We demonstrate the performance of our proposed β-divergence-based autoencoder for a variety of image and categorical datasets showing improved robustness to outliers both qualitatively and quantitatively. We also illustrate the use of our robust VAE for detection of lesions in brain images, formulated as an anomaly detection task. Finally, we suggest a method to tune the hyperparameter of RVAE which makes our model completely unsupervised. |
| ArticleNumber | 107886 |
| Author | Joshi, Anand A. Aydöre, Sergül Leahy, Richard M. Akrami, Haleh Li, Jian |
| Author_xml | – sequence: 1 givenname: Haleh orcidid: 0000-0002-1678-8926 surname: Akrami fullname: Akrami, Haleh email: akrami@usc.edu organization: Signal and Image Processing Institute, University of Southern California, Los Angeles, CA, USA – sequence: 2 givenname: Anand A. orcidid: 0000-0002-9582-3848 surname: Joshi fullname: Joshi, Anand A. email: ajoshi@usc.edu organization: Signal and Image Processing Institute, University of Southern California, Los Angeles, CA, USA – sequence: 3 givenname: Jian orcidid: 0000-0002-1691-8727 surname: Li fullname: Li, Jian email: jli112@mgh.harvard.edu organization: Athinoula A. Martinos Center for Biomedical Imaging Massachusetts General Hospital, Harvard Medical School, Charlestown, MA, USA – sequence: 4 givenname: Sergül surname: Aydöre fullname: Aydöre, Sergül email: sergulaydore@gmail.com organization: Amazon Web Services, New York, NY, USA – sequence: 5 givenname: Richard M. orcidid: 0000-0002-7278-5471 surname: Leahy fullname: Leahy, Richard M. email: leahy@sipi.usc.edu organization: Signal and Image Processing Institute, University of Southern California, Los Angeles, CA, USA |
| BackLink | https://www.ncbi.nlm.nih.gov/pubmed/36714396$$D View this record in MEDLINE/PubMed |
| BookMark | eNqFkE1LHTEUhkOx1KvtP5Ay4Kabuc3X5MOFcBH7AUI3ug6Z5IzkOndik8wF_31zHd24sKsD5zzvC-c5QUdTnAChM4LXBBPxfbt-mGJ-ymuKKakrqZT4gFZESdpKjvURWmHd4Vbijhyjk5y3GGNKifqEjpmQhDMtVkhsmhT7OZdmb1OwJcTJjo2dS4TJRQ-pmXOY7pseim182EO6rwf4jD4Odszw5WWeorsf17dXv9qbPz9_X21uWseZKK3quHAD99rD0IPUVPe8IyCsEz23vKNSOEYtiK5nWhNusbZkIH6QomOEenaKvi29jyn-nSEXswvZwTjaCeKcDZWSYKWZkhU9f4Nu45zqN5USTGFFdCcq9fWFmvsdePOYws6mJ_NqpAIXC-BSzDnBYFwoz15KsmE0BJuDfrM1i35z0G8W_TXM34Rf-_8Tu1xiUFXuAySTXTho9iGBK8bH8H7BPwoQn24 |
| CitedBy_id | crossref_primary_10_1016_j_eswa_2023_121074 crossref_primary_10_1016_j_automatica_2024_112108 crossref_primary_10_1016_j_cose_2023_103251 crossref_primary_10_26599_CVM_2025_9450403 crossref_primary_10_1016_j_engappai_2023_106684 crossref_primary_10_1109_ACCESS_2025_3594877 crossref_primary_10_1007_s11368_024_03801_1 crossref_primary_10_1016_j_future_2024_107630 crossref_primary_10_1109_TASE_2024_3486688 crossref_primary_10_3390_app15115830 crossref_primary_10_1051_epjconf_202429509033 crossref_primary_10_1016_j_eswa_2023_120214 crossref_primary_10_1016_j_ascom_2023_100739 crossref_primary_10_1016_j_neucom_2025_131423 crossref_primary_10_1007_s00500_025_10702_z crossref_primary_10_1016_j_knosys_2023_110287 crossref_primary_10_2196_77893 crossref_primary_10_1016_j_chemolab_2024_105276 crossref_primary_10_1002_hbm_70075 crossref_primary_10_1016_j_swevo_2024_101520 crossref_primary_10_1109_TIA_2025_3549413 |
| Cites_doi | 10.1126/science.1127647 10.1080/00031305.1988.10475585 10.1080/03610928808829834 10.3390/e12061532 10.1016/j.media.2016.07.009 10.1109/5.726791 10.1016/j.media.2020.101713 10.1109/MCSE.2011.37 10.1109/ICCV.2019.01037 10.1093/biomet/85.3.549 10.3390/e12020262 10.1093/neuros/nyx103 10.1080/00949659408811609 10.1609/aaai.v31i1.10777 10.1016/0377-0427(87)90125-7 |
| ContentType | Journal Article |
| Copyright | 2021 Elsevier B.V. Copyright Elsevier Science Ltd. Feb 28, 2022 |
| Copyright_xml | – notice: 2021 Elsevier B.V. – notice: Copyright Elsevier Science Ltd. Feb 28, 2022 |
| DBID | AAYXX CITATION NPM 7SC 8FD E3H F2A JQ2 L7M L~C L~D 7X8 |
| DOI | 10.1016/j.knosys.2021.107886 |
| DatabaseName | CrossRef PubMed Computer and Information Systems Abstracts Technology Research Database Library & Information Sciences Abstracts (LISA) Library & Information Science Abstracts (LISA) ProQuest Computer Science Collection Advanced Technologies Database with Aerospace Computer and Information Systems Abstracts Academic Computer and Information Systems Abstracts Professional MEDLINE - Academic |
| DatabaseTitle | CrossRef PubMed Technology Research Database Computer and Information Systems Abstracts – Academic Library and Information Science Abstracts (LISA) ProQuest Computer Science Collection Computer and Information Systems Abstracts Advanced Technologies Database with Aerospace Computer and Information Systems Abstracts Professional MEDLINE - Academic |
| DatabaseTitleList | PubMed MEDLINE - Academic Technology Research Database |
| Database_xml | – sequence: 1 dbid: NPM name: PubMed url: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed sourceTypes: Index Database – sequence: 2 dbid: 7X8 name: MEDLINE - Academic url: https://search.proquest.com/medline sourceTypes: Aggregation Database |
| DeliveryMethod | fulltext_linktorsrc |
| Discipline | Computer Science Statistics |
| EISSN | 1872-7409 |
| ExternalDocumentID | 36714396 10_1016_j_knosys_2021_107886 S0950705121010534 |
| Genre | Journal Article |
| GrantInformation_xml | – fundername: NINDS NIH HHS grantid: R01 NS074980 – fundername: NIBIB NIH HHS grantid: R01 EB026299 |
| GroupedDBID | --K --M .DC .~1 0R~ 1B1 1~. 1~5 4.4 457 4G. 5VS 7-5 71M 77K 8P~ 9JN AACTN AAEDT AAEDW AAIAV AAIKJ AAKOC AALRI AAOAW AAQFI AAXUO AAYFN ABAOU ABBOA ABIVO ABJNI ABMAC ABYKQ ACAZW ACDAQ ACGFS ACRLP ACZNC ADBBV ADEZE ADGUI ADTZH AEBSH AECPX AEKER AENEX AFKWA AFTJW AGHFR AGUBO AGYEJ AHHHB AHJVU AHZHX AIALX AIEXJ AIKHN AITUG AJOXV ALMA_UNASSIGNED_HOLDINGS AMFUW AMRAJ AOUOD ARUGR AXJTR BJAXD BKOJK BLXMC CS3 DU5 EBS EFJIC EFLBG EO8 EO9 EP2 EP3 FDB FIRID FNPLU FYGXN G-Q GBLVA GBOLZ IHE J1W JJJVA KOM LG9 LY7 M41 MHUIS MO0 N9A O-L O9- OAUVE OZT P-8 P-9 P2P PC. PQQKQ Q38 ROL RPZ SDF SDG SDP SES SPC SPCBC SST SSV SSW SSZ T5K WH7 XPP ZMT ~02 ~G- 29L 77I 9DU AAQXK AATTM AAXKI AAYWO AAYXX ABDPE ABWVN ABXDB ACLOT ACNNM ACRPL ACVFH ADCNI ADJOM ADMUD ADNMO AEIPS AEUPX AFJKZ AFPUW AGQPQ AIGII AIIUN AKBMS AKRWK AKYEP ANKPU APXCP ASPBG AVWKF AZFZN CITATION EFKBS EJD FEDTE FGOYB G-2 HLZ HVGLF HZ~ R2- SBC SET SEW UHS WUQ ~HD AFXIZ AGCQF AGRNS BNPGV NPM SSH 7SC 8FD E3H F2A JQ2 L7M L~C L~D 7X8 |
| ID | FETCH-LOGICAL-c436t-8546cf4d9defbe7929b451e6ac6b4a45276c32ae65b39914a09a1f1df765312d3 |
| ISICitedReferencesCount | 31 |
| ISICitedReferencesURI | http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000779159800013&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| ISSN | 0950-7051 |
| IngestDate | Sun Sep 28 01:45:17 EDT 2025 Fri Nov 14 18:45:26 EST 2025 Mon Jul 21 06:04:55 EDT 2025 Tue Nov 18 21:55:43 EST 2025 Sat Nov 29 07:07:00 EST 2025 Fri Feb 23 02:41:34 EST 2024 |
| IsDoiOpenAccess | false |
| IsOpenAccess | true |
| IsPeerReviewed | true |
| IsScholarly | true |
| Keywords | Outlier β divergence Robust anomaly detection VAE RVAE β divergence |
| Language | English |
| LinkModel | OpenURL |
| MergedId | FETCHMERGED-LOGICAL-c436t-8546cf4d9defbe7929b451e6ac6b4a45276c32ae65b39914a09a1f1df765312d3 |
| Notes | ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14 content type line 23 |
| ORCID | 0000-0002-9582-3848 0000-0002-1678-8926 0000-0002-1691-8727 0000-0002-7278-5471 |
| OpenAccessLink | https://www.ncbi.nlm.nih.gov/pmc/articles/9881733 |
| PMID | 36714396 |
| PQID | 2638081956 |
| PQPubID | 2035257 |
| ParticipantIDs | proquest_miscellaneous_2771089387 proquest_journals_2638081956 pubmed_primary_36714396 crossref_citationtrail_10_1016_j_knosys_2021_107886 crossref_primary_10_1016_j_knosys_2021_107886 elsevier_sciencedirect_doi_10_1016_j_knosys_2021_107886 |
| PublicationCentury | 2000 |
| PublicationDate | 2022-02-28 |
| PublicationDateYYYYMMDD | 2022-02-28 |
| PublicationDate_xml | – month: 02 year: 2022 text: 2022-02-28 day: 28 |
| PublicationDecade | 2020 |
| PublicationPlace | Netherlands |
| PublicationPlace_xml | – name: Netherlands – name: Amsterdam |
| PublicationTitle | Knowledge-based systems |
| PublicationTitleAlternate | Knowl Based Syst |
| PublicationYear | 2022 |
| Publisher | Elsevier B.V Elsevier Science Ltd |
| Publisher_xml | – name: Elsevier B.V – name: Elsevier Science Ltd |
| References | Larsen, Sønderby, Larochelle, Winther (b39) 2015 Pawlowski, Lee, Rajchl, McDonagh, Ferrante, Kamnitsas, Cooke, Stevenson, Khetani, Newman (b13) 2018 Chen, You, Tezcan, Konukoglu (b15) 2020 An, Cho (b10) 2015 Zhou, Paffenroth (b20) 2017 Kusner (b7) 2017 Vincent (b18) 2008 Walt, Colbert, Varoquaux (b42) 2011; 13 D. Im Im, S. Ahn, R. Memisevic, Y. Bengio, Denoising criterion for variational auto-encoding framework, in: Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 31, 2017. Kingma, Ba (b43) 2014 Basu, Harris, Hjort, Jones (b23) 1998; 85 Cichocki, Amari (b28) 2010; 12 X. Ma, A.R. Triki, M. Berman, C. Sagonas, J. Cali, M.B. Blaschko, A Bayesian optimization framework for neural network compression, in: Proceedings of the IEEE International Conference on Computer Vision, 2019, pp. 10274–10283. (b36) 2020 Hsu, Zhang, Glass (b9) 2017 Eguchi, Kato (b24) 2010; 12 Brent (b45) 2013 LeCun, Bottou, Bengio, Haffner (b30) 1998; 86 (b34) 1999 Pu, Gan, Henao, Yuan, Li, Stevens, Carin (b8) 2016 Loaiza-Ganem, Cunningham (b44) 2019 Gather, Kale (b2) 1988; 17 (b35) 2020 Qi, Wang, Zheng, Wu (b19) 2014 Hinton, Salakhutdinov (b1) 2006; 313 Baur, Wiestler, Albarqouni, Navab (b12) 2018 Wingate, Weber (b29) 2013 Cao, Li, Nelson, Kon (b22) 2019 You, Tezcan, Chen, Konukoglu (b11) 2019 Eduardo (b17) 2019 Futami, Sato, Sugiyama (b25) 2017 Goodfellow, Pouget-Abadie, Mirza, Xu, Warde-Farley, Ozair, Courville, Bengio (b53) 2014 Cohen, Afshar, Tapson, van Schaik (b31) 2017 Maier (b33) 2017; 35 Kingma, Welling (b5) 2013 Nalisnick, Matsukawa, Teh, Gorur, Lakshminarayanan (b16) 2018 Villanueva-Meyer, Mabray, Cha (b37) 2017; 81 Xiao, Rasul, Vollgraf (b32) 2017 Paszke, Gross, Chintala, Chanan, Yang, DeVito, Lin, Desmaison, Antiga, Lerer (b40) 2017 Rousseeuw (b49) 1987; 20 Zimmerer, Kohl, Petersen, Isensee, Maier-Hein (b14) 2018 Zellner (b27) 1988; 42 Chen, Konukoglu (b50) 2018 Press, Flannery, Teukolsky, Vetterling (b46) 1989 Bishop (b6) 2006 Dewancker, McCourt, Clark (b47) 2016 Basu, Sarkar (b51) 1994; 50 Dai, Wipf (b52) 2019 Dai, Wang, Aston, Hua, Wipf (b26) 2018; 19 Huber (b3) 2011 Zhai, Chen, Zhang, Wang (b21) 2017 Hampel, Ronchetti, Rousseeuw, Stahel (b4) 1986 Pedregosa, Varoquaux, Gramfort, Michel, Thirion, Grisel, Blondel, Prettenhofer, Weiss, Dubourg (b41) 2011; 12 Dewancker (10.1016/j.knosys.2021.107886_b47) 2016 Loaiza-Ganem (10.1016/j.knosys.2021.107886_b44) 2019 Rousseeuw (10.1016/j.knosys.2021.107886_b49) 1987; 20 You (10.1016/j.knosys.2021.107886_b11) 2019 Kusner (10.1016/j.knosys.2021.107886_b7) 2017 Zhai (10.1016/j.knosys.2021.107886_b21) 2017 Cohen (10.1016/j.knosys.2021.107886_b31) 2017 Walt (10.1016/j.knosys.2021.107886_b42) 2011; 13 10.1016/j.knosys.2021.107886_b38 Huber (10.1016/j.knosys.2021.107886_b3) 2011 Paszke (10.1016/j.knosys.2021.107886_b40) 2017 Hsu (10.1016/j.knosys.2021.107886_b9) 2017 Kingma (10.1016/j.knosys.2021.107886_b43) 2014 (10.1016/j.knosys.2021.107886_b34) 1999 Baur (10.1016/j.knosys.2021.107886_b12) 2018 Chen (10.1016/j.knosys.2021.107886_b15) 2020 (10.1016/j.knosys.2021.107886_b36) 2020 Larsen (10.1016/j.knosys.2021.107886_b39) 2015 Dai (10.1016/j.knosys.2021.107886_b26) 2018; 19 Pedregosa (10.1016/j.knosys.2021.107886_b41) 2011; 12 Gather (10.1016/j.knosys.2021.107886_b2) 1988; 17 LeCun (10.1016/j.knosys.2021.107886_b30) 1998; 86 Villanueva-Meyer (10.1016/j.knosys.2021.107886_b37) 2017; 81 Basu (10.1016/j.knosys.2021.107886_b51) 1994; 50 Goodfellow (10.1016/j.knosys.2021.107886_b53) 2014 Bishop (10.1016/j.knosys.2021.107886_b6) 2006 Eduardo (10.1016/j.knosys.2021.107886_b17) 2019 Press (10.1016/j.knosys.2021.107886_b46) 1989 Hinton (10.1016/j.knosys.2021.107886_b1) 2006; 313 Cichocki (10.1016/j.knosys.2021.107886_b28) 2010; 12 (10.1016/j.knosys.2021.107886_b35) 2020 Maier (10.1016/j.knosys.2021.107886_b33) 2017; 35 Zellner (10.1016/j.knosys.2021.107886_b27) 1988; 42 An (10.1016/j.knosys.2021.107886_b10) 2015 Qi (10.1016/j.knosys.2021.107886_b19) 2014 Eguchi (10.1016/j.knosys.2021.107886_b24) 2010; 12 Hampel (10.1016/j.knosys.2021.107886_b4) 1986 Futami (10.1016/j.knosys.2021.107886_b25) 2017 Brent (10.1016/j.knosys.2021.107886_b45) 2013 Basu (10.1016/j.knosys.2021.107886_b23) 1998; 85 Zimmerer (10.1016/j.knosys.2021.107886_b14) 2018 Zhou (10.1016/j.knosys.2021.107886_b20) 2017 Wingate (10.1016/j.knosys.2021.107886_b29) 2013 Kingma (10.1016/j.knosys.2021.107886_b5) 2013 Chen (10.1016/j.knosys.2021.107886_b50) 2018 Dai (10.1016/j.knosys.2021.107886_b52) 2019 Pawlowski (10.1016/j.knosys.2021.107886_b13) 2018 Pu (10.1016/j.knosys.2021.107886_b8) 2016 Cao (10.1016/j.knosys.2021.107886_b22) 2019 Nalisnick (10.1016/j.knosys.2021.107886_b16) 2018 Xiao (10.1016/j.knosys.2021.107886_b32) 2017 10.1016/j.knosys.2021.107886_b48 Vincent (10.1016/j.knosys.2021.107886_b18) 2008 |
| References_xml | – volume: 85 start-page: 549 year: 1998 end-page: 559 ident: b23 article-title: Robust and efficient estimation by minimising a density power divergence publication-title: Biometrika – year: 1986 ident: b4 article-title: Robust Statistics – year: 2018 ident: b50 article-title: Unsupervised detection of lesions in brain MRI using constrained adversarial auto-encoders – start-page: 1945 year: 2017 end-page: 1954 ident: b7 article-title: Grammar variational autoencoder publication-title: Proceedings of the 34th International Conference on Machine Learning-Volume 70 – year: 2018 ident: b13 article-title: Unsupervised lesion detection in brain CT using bayesian convolutional autoencoders publication-title: MIDL, Abstract Track, Non-Archival – volume: 50 start-page: 173 year: 1994 end-page: 185 ident: b51 article-title: The trade-off between robustness and efficiency and the effect of model smoothing in minimum disparity inference publication-title: J. Stat. Comput. Simul. – volume: 81 start-page: 397 year: 2017 end-page: 415 ident: b37 article-title: Current clinical brain tumor imaging publication-title: Neurosurgery – year: 2020 ident: b15 article-title: Unsupervised lesion detection via image restoration with a normative prior publication-title: Med. Image Anal. – volume: 12 start-page: 262 year: 2010 end-page: 274 ident: b24 article-title: Entropy and divergence associated with power function and the statistical application publication-title: Entropy – year: 2006 ident: b6 article-title: Pattern Recognition and Machine Learning – start-page: 161 year: 2018 end-page: 169 ident: b12 article-title: Deep autoencoding models for unsupervised anomaly segmentation in brain mr images publication-title: International MICCAI Brainlesion Workshop – start-page: 13266 year: 2019 end-page: 13276 ident: b44 article-title: The continuous Bernoulli: fixing a pervasive error in variational autoencoders publication-title: Advances in Neural Information Processing Systems – year: 2015 ident: b39 article-title: Autoencoding beyond pixels using a learned similarity metric – volume: 12 start-page: 2825 year: 2011 end-page: 2830 ident: b41 article-title: Scikit-learn: Machine learning in Python publication-title: J. Mach. Learn. Res. – year: 2014 ident: b43 article-title: Adam: A method for stochastic optimization – volume: 86 start-page: 2278 year: 1998 end-page: 2324 ident: b30 article-title: Gradient-based learning applied to document recognition publication-title: Proc. IEEE – year: 2013 ident: b29 article-title: Automated variational inference in probabilistic programming – year: 2013 ident: b45 article-title: Algorithms for Minimization Without Derivatives – volume: 12 start-page: 1532 year: 2010 end-page: 1568 ident: b28 article-title: Families of alpha-beta-and gamma-divergences: Flexible and robust measures of similarities publication-title: Entropy – year: 1999 ident: b34 article-title: KDD cup 1999 data – volume: 313 start-page: 504 year: 2006 end-page: 507 ident: b1 article-title: Reducing the dimensionality of data with neural networks publication-title: Science – start-page: 6716 year: 2014 end-page: 6720 ident: b19 article-title: Robust feature learning by stacked autoencoder with maximum correntropy criterion publication-title: 2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) – start-page: 540 year: 2019 end-page: 556 ident: b11 article-title: Unsupervised lesion detection via image restoration with a normative prior publication-title: International Conference on Medical Imaging with Deep Learning – year: 2019 ident: b22 article-title: Coupled VAE: Improved accuracy and robustness of a variational autoencoder – year: 1989 ident: b46 article-title: Numerical Recipes, Vol. 3 – start-page: 2352 year: 2016 end-page: 2360 ident: b8 article-title: Variational autoencoder for deep learning of images, labels and captions publication-title: Advances in Neural Information Processing Systems – year: 2016 ident: b47 article-title: Bayesian optimization for machine learning: A practical guidebook – volume: 13 start-page: 22 year: 2011 end-page: 30 ident: b42 article-title: The NumPy array: a structure for efficient numerical computation publication-title: Comput. Sci. Eng. – start-page: 2672 year: 2014 end-page: 2680 ident: b53 article-title: Generative adversarial nets publication-title: Advances in Neural Information Processing Systems – year: 2019 ident: b17 article-title: Robust variational autoencoders for outlier detection in mixed-type data – year: 2017 ident: b9 article-title: Learning latent representations for speech generation and transformation – start-page: 356 year: 2017 end-page: 367 ident: b21 article-title: Robust variational auto-encoder for radar HRRP target recognition publication-title: International Conference on Intelligent Science and Big Data Engineering – volume: 20 start-page: 53 year: 1987 end-page: 65 ident: b49 article-title: Silhouettes: a graphical aid to the interpretation and validation of cluster analysis publication-title: J. Comput. Appl. Math. – year: 2017 ident: b25 article-title: Variational inference based on robust divergences – volume: 42 start-page: 278 year: 1988 end-page: 280 ident: b27 article-title: Optimal information processing and Bayes’s theorem publication-title: Amer. Statist. – year: 2013 ident: b5 article-title: Auto-encoding variational Bayes – reference: X. Ma, A.R. Triki, M. Berman, C. Sagonas, J. Cali, M.B. Blaschko, A Bayesian optimization framework for neural network compression, in: Proceedings of the IEEE International Conference on Computer Vision, 2019, pp. 10274–10283. – volume: 19 start-page: 1573 year: 2018 end-page: 1614 ident: b26 article-title: Connections with robust PCA and the role of emergent sparsity in variational autoencoder models publication-title: J. Mach. Learn. Res. – volume: 35 start-page: 250 year: 2017 end-page: 269 ident: b33 article-title: ISLES 2015-A public evaluation benchmark for ischemic stroke lesion segmentation from multispectral MRI publication-title: Med. Image Anal. – year: 2020 ident: b35 article-title: NSL-KDD dataset – start-page: 665 year: 2017 end-page: 674 ident: b20 article-title: Anomaly detection with robust deep autoencoders publication-title: Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining – start-page: 1 year: 2015 end-page: 18 ident: b10 article-title: Variational autoencoder based anomaly detection using reconstruction probability publication-title: Special Lecture on IE, Vol. 2 – year: 2018 ident: b16 article-title: Do deep generative models know what they don’t know? – year: 2017 ident: b32 article-title: Fashion-MNIST: a novel image dataset for benchmarking machine learning algorithms – start-page: 1096 year: 2008 end-page: 1103 ident: b18 article-title: Extracting and composing robust features with denoising autoencoders publication-title: Proceedings of the 25th International Conference on Machine Learning – volume: 17 start-page: 3767 year: 1988 end-page: 3784 ident: b2 article-title: Maximum likelihood estimation in the presence of outiliers publication-title: Comm. Statist. Theory Methods – year: 2018 ident: b14 article-title: Context-encoding variational autoencoder for unsupervised anomaly detection – year: 2017 ident: b31 article-title: EMNIST: an extension of MNIST to handwritten letters – year: 2017 ident: b40 article-title: Automatic differentiation in pytorch publication-title: NIPS-W – year: 2019 ident: b52 article-title: Diagnosing and enhancing VAE models – year: 2020 ident: b36 article-title: UNSW-NB15 – reference: D. Im Im, S. Ahn, R. Memisevic, Y. Bengio, Denoising criterion for variational auto-encoding framework, in: Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 31, 2017. – year: 2011 ident: b3 article-title: Robust Statistics – start-page: 540 year: 2019 ident: 10.1016/j.knosys.2021.107886_b11 article-title: Unsupervised lesion detection via image restoration with a normative prior – year: 2017 ident: 10.1016/j.knosys.2021.107886_b40 article-title: Automatic differentiation in pytorch – volume: 313 start-page: 504 issue: 5786 year: 2006 ident: 10.1016/j.knosys.2021.107886_b1 article-title: Reducing the dimensionality of data with neural networks publication-title: Science doi: 10.1126/science.1127647 – volume: 42 start-page: 278 issue: 4 year: 1988 ident: 10.1016/j.knosys.2021.107886_b27 article-title: Optimal information processing and Bayes’s theorem publication-title: Amer. Statist. doi: 10.1080/00031305.1988.10475585 – year: 2017 ident: 10.1016/j.knosys.2021.107886_b32 – year: 2016 ident: 10.1016/j.knosys.2021.107886_b47 – year: 2013 ident: 10.1016/j.knosys.2021.107886_b5 – start-page: 356 year: 2017 ident: 10.1016/j.knosys.2021.107886_b21 article-title: Robust variational auto-encoder for radar HRRP target recognition – start-page: 13266 year: 2019 ident: 10.1016/j.knosys.2021.107886_b44 article-title: The continuous Bernoulli: fixing a pervasive error in variational autoencoders – volume: 17 start-page: 3767 issue: 11 year: 1988 ident: 10.1016/j.knosys.2021.107886_b2 article-title: Maximum likelihood estimation in the presence of outiliers publication-title: Comm. Statist. Theory Methods doi: 10.1080/03610928808829834 – year: 2013 ident: 10.1016/j.knosys.2021.107886_b45 – year: 2019 ident: 10.1016/j.knosys.2021.107886_b52 – start-page: 1 year: 2015 ident: 10.1016/j.knosys.2021.107886_b10 article-title: Variational autoencoder based anomaly detection using reconstruction probability – volume: 12 start-page: 1532 issue: 6 year: 2010 ident: 10.1016/j.knosys.2021.107886_b28 article-title: Families of alpha-beta-and gamma-divergences: Flexible and robust measures of similarities publication-title: Entropy doi: 10.3390/e12061532 – year: 1999 ident: 10.1016/j.knosys.2021.107886_b34 – year: 2020 ident: 10.1016/j.knosys.2021.107886_b35 – volume: 35 start-page: 250 year: 2017 ident: 10.1016/j.knosys.2021.107886_b33 article-title: ISLES 2015-A public evaluation benchmark for ischemic stroke lesion segmentation from multispectral MRI publication-title: Med. Image Anal. doi: 10.1016/j.media.2016.07.009 – year: 2020 ident: 10.1016/j.knosys.2021.107886_b36 – year: 2018 ident: 10.1016/j.knosys.2021.107886_b50 – year: 2015 ident: 10.1016/j.knosys.2021.107886_b39 – year: 2017 ident: 10.1016/j.knosys.2021.107886_b25 – start-page: 161 year: 2018 ident: 10.1016/j.knosys.2021.107886_b12 article-title: Deep autoencoding models for unsupervised anomaly segmentation in brain mr images – year: 2019 ident: 10.1016/j.knosys.2021.107886_b17 – volume: 19 start-page: 1573 issue: 1 year: 2018 ident: 10.1016/j.knosys.2021.107886_b26 article-title: Connections with robust PCA and the role of emergent sparsity in variational autoencoder models publication-title: J. Mach. Learn. Res. – year: 1986 ident: 10.1016/j.knosys.2021.107886_b4 – year: 2018 ident: 10.1016/j.knosys.2021.107886_b16 – start-page: 2672 year: 2014 ident: 10.1016/j.knosys.2021.107886_b53 article-title: Generative adversarial nets – volume: 86 start-page: 2278 issue: 11 year: 1998 ident: 10.1016/j.knosys.2021.107886_b30 article-title: Gradient-based learning applied to document recognition publication-title: Proc. IEEE doi: 10.1109/5.726791 – year: 2020 ident: 10.1016/j.knosys.2021.107886_b15 article-title: Unsupervised lesion detection via image restoration with a normative prior publication-title: Med. Image Anal. doi: 10.1016/j.media.2020.101713 – year: 1989 ident: 10.1016/j.knosys.2021.107886_b46 – volume: 13 start-page: 22 issue: 2 year: 2011 ident: 10.1016/j.knosys.2021.107886_b42 article-title: The NumPy array: a structure for efficient numerical computation publication-title: Comput. Sci. Eng. doi: 10.1109/MCSE.2011.37 – ident: 10.1016/j.knosys.2021.107886_b48 doi: 10.1109/ICCV.2019.01037 – start-page: 2352 year: 2016 ident: 10.1016/j.knosys.2021.107886_b8 article-title: Variational autoencoder for deep learning of images, labels and captions – year: 2014 ident: 10.1016/j.knosys.2021.107886_b43 – year: 2018 ident: 10.1016/j.knosys.2021.107886_b13 article-title: Unsupervised lesion detection in brain CT using bayesian convolutional autoencoders – volume: 85 start-page: 549 issue: 3 year: 1998 ident: 10.1016/j.knosys.2021.107886_b23 article-title: Robust and efficient estimation by minimising a density power divergence publication-title: Biometrika doi: 10.1093/biomet/85.3.549 – volume: 12 start-page: 262 issue: 2 year: 2010 ident: 10.1016/j.knosys.2021.107886_b24 article-title: Entropy and divergence associated with power function and the statistical application publication-title: Entropy doi: 10.3390/e12020262 – year: 2017 ident: 10.1016/j.knosys.2021.107886_b9 – year: 2019 ident: 10.1016/j.knosys.2021.107886_b22 – volume: 81 start-page: 397 issue: 3 year: 2017 ident: 10.1016/j.knosys.2021.107886_b37 article-title: Current clinical brain tumor imaging publication-title: Neurosurgery doi: 10.1093/neuros/nyx103 – year: 2006 ident: 10.1016/j.knosys.2021.107886_b6 – year: 2018 ident: 10.1016/j.knosys.2021.107886_b14 – volume: 50 start-page: 173 issue: 3–4 year: 1994 ident: 10.1016/j.knosys.2021.107886_b51 article-title: The trade-off between robustness and efficiency and the effect of model smoothing in minimum disparity inference publication-title: J. Stat. Comput. Simul. doi: 10.1080/00949659408811609 – start-page: 1096 year: 2008 ident: 10.1016/j.knosys.2021.107886_b18 article-title: Extracting and composing robust features with denoising autoencoders – start-page: 665 year: 2017 ident: 10.1016/j.knosys.2021.107886_b20 article-title: Anomaly detection with robust deep autoencoders – ident: 10.1016/j.knosys.2021.107886_b38 doi: 10.1609/aaai.v31i1.10777 – year: 2017 ident: 10.1016/j.knosys.2021.107886_b31 – start-page: 6716 year: 2014 ident: 10.1016/j.knosys.2021.107886_b19 article-title: Robust feature learning by stacked autoencoder with maximum correntropy criterion – start-page: 1945 year: 2017 ident: 10.1016/j.knosys.2021.107886_b7 article-title: Grammar variational autoencoder – volume: 12 start-page: 2825 issue: Oct year: 2011 ident: 10.1016/j.knosys.2021.107886_b41 article-title: Scikit-learn: Machine learning in Python publication-title: J. Mach. Learn. Res. – volume: 20 start-page: 53 year: 1987 ident: 10.1016/j.knosys.2021.107886_b49 article-title: Silhouettes: a graphical aid to the interpretation and validation of cluster analysis publication-title: J. Comput. Appl. Math. doi: 10.1016/0377-0427(87)90125-7 – year: 2011 ident: 10.1016/j.knosys.2021.107886_b3 – year: 2013 ident: 10.1016/j.knosys.2021.107886_b29 |
| SSID | ssj0002218 |
| Score | 2.5187771 |
| Snippet | The presence of outliers can severely degrade learned representations and performance of deep learning methods and hence disproportionately affect the training... |
| SourceID | proquest pubmed crossref elsevier |
| SourceType | Aggregation Database Index Database Enrichment Source Publisher |
| StartPage | 107886 |
| SubjectTerms | Anomalies Brain damage Data Data analysis Deep learning Lesions Lower bounds Neuroimaging Outlier Outliers (statistics) Robust anomaly detection Robust control Robustness RVAE Statistics Training VAE β divergence |
| Title | A robust variational autoencoder using beta divergence |
| URI | https://dx.doi.org/10.1016/j.knosys.2021.107886 https://www.ncbi.nlm.nih.gov/pubmed/36714396 https://www.proquest.com/docview/2638081956 https://www.proquest.com/docview/2771089387 |
| Volume | 238 |
| WOSCitedRecordID | wos000779159800013&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: PRVESC databaseName: Elsevier SD Freedom Collection Journals 2021 customDbUrl: eissn: 1872-7409 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0002218 issn: 0950-7051 databaseCode: AIEXJ dateStart: 19950201 isFulltext: true titleUrlDefault: https://www.sciencedirect.com providerName: Elsevier |
| link | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwtV3da9swEBdpO0Zf9v2RrSse7C14xLYsWS8Dr3RsWSl76EbejCzJsDa1Q2KH7r_f6csxK6Hbw16MLclG3J3ufneW7hB6l0U8TkuahiVY_xCraRLqkSFhpawqnBBm0hf_OKPn59l8zr6NRh_8WZjNgtZ1dnPDlv-V1dAGzNZHZ_-B3f1HoQHugelwBbbD9a8Yn09WTdmt28kG3GAf6uNd2-iUlTpzRGfCA6Vq-UTqXRkmHecQpH71cbZQ2zjpsj334Du_WvFrW-warEsfTp41a1MgWOc5qeU2RnpmGmcDMcx_Sf1__iNx8W-9w0w_nyyGMQhwX7dnun0wcRrSqUsd6_RqbNO23NLRNlxw-f6qbmD-4KLHETSCK06Gw4Goy2vDooToGu3sj4TZ1gS7rj10ENOUgVY7yL-czme9LY5jE-Ht5-cPT5odfrcncIju-0_uwim7_BCDRy4eoQfOkQhyKwCP0UjVT9BDX6QjcDr7KSJ5YOUhGMhDMJCHwMhDoOUh2MrDM_T90-nFyefQFcsIBayoNsxSTESFJZOqKhUF1FviNFKEC1JijtOYEpHEXJG0BEwaYT5lPKoiWVECajiWyXO0Xze1eokCToVkkeIZFykWuOSSJxLc2qzCTLFKjFHiSVMIl0leFzRZFH7L4GVhaVto2haWtmMU9m8tbSaVO8ZTT_XCoUGL8goQpDvePPJMKtzChH4wNBr-ptD9tu8GXap_kPFaNR2MoYC3AcBndIxeWOb2U_Vy8Wpnz2t0uF0bR2i_XXXqDbonNu3P9eoY7dF5duzk8zdzDJZR |
| linkProvider | Elsevier |
| 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=A+robust+variational+autoencoder+using+beta+divergence&rft.jtitle=Knowledge-based+systems&rft.au=Akrami%2C+Haleh&rft.au=Joshi%2C+Anand+A&rft.au=Li%2C+Jian&rft.au=Ayd%C3%B6re%2C+Serg%C3%BCl&rft.date=2022-02-28&rft.issn=0950-7051&rft.volume=238&rft_id=info:doi/10.1016%2Fj.knosys.2021.107886&rft_id=info%3Apmid%2F36714396&rft.externalDocID=36714396 |
| thumbnail_l | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=0950-7051&client=summon |
| thumbnail_m | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=0950-7051&client=summon |
| thumbnail_s | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=0950-7051&client=summon |