Synthesizing data using variational autoencoders for handling class imbalanced deep learning

Taimoor Shakeel Sheikh; Adil Khan; Muhammad Fahim; Muhammad Ahmad

doi:10.1007/978-3-030-39575-9_28

Synthesizing data using variational autoencoders for handling class imbalanced deep learning

Taimoor Shakeel Sheikh^*, Adil Khan, Muhammad Fahim, Muhammad Ahmad

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

This paper addresses the complex problem of learning from unbalanced datasets due to which traditional algorithms may perform poorly. Classification algorithms used for learning tend to favor the larger, less important classes in such problems. In this work, to handle unbalanced data problem, we synthesize data using variational autoencoders (VAE) on raw training samples and then, use various input sources (raw, combination of raw and synthetic) to train different models. We evaluate our method using multiple criteria on SVHN dataset which consists of complex images, and perform a comprehensive comparative analysis of popular CNN architectures when there is balanced and unbalanced data and determine which operates best in class imbalance problem. We found that data synthesis via VAE is reliable and robust, and can help to classify real data with higher accuracy than traditional (unbalanced) data. Our results demonstrate the strength of using VAE to solve the class imbalance problem.

Original language	English
Title of host publication	Analysis of Images, Social Networks and Texts - 8th International Conference, AIST 2019, Revised Selected Papers: Proceedings
Editors	Wil M.P. van der Aalst, Vladimir Batagelj, Dmitry I. Ignatov, Valentina Kuskova, Sergei O. Kuznetsov, Irina A. Lomazova, Michael Khachay, Andrey Kutuzov, Natalia Loukachevitch, Amedeo Napoli, Panos M. Pardalos, Marcello Pelillo, Andrey V. Savchenko, Elena Tutubalina
Publisher	Springer
Pages	270-281
Number of pages	12
ISBN (Print)	9783030395742
DOIs	https://doi.org/10.1007/978-3-030-39575-9_28
Publication status	Published - 02 Feb 2020
Externally published	Yes
Event	8th International Conference on Analysis of Images, Social Networks and Texts, AIST 2019 - Kazan, Russian Federation Duration: 17 Jul 2019 → 19 Jul 2019

Publication series

Name	Communications in Computer and Information Science
Volume	1086CCIS
ISSN (Print)	1865-0929
ISSN (Electronic)	1865-0937

Conference

Conference	8th International Conference on Analysis of Images, Social Networks and Texts, AIST 2019
Country/Territory	Russian Federation
City	Kazan
Period	17/07/2019 → 19/07/2019

Bibliographical note

Publisher Copyright:
© Springer Nature Switzerland AG 2020.

Copyright:
Copyright 2020 Elsevier B.V., All rights reserved.

Keywords

Convolutional Neural Network (CNN)
Imbalanced data
Variational autoencoder (VAE)

ASJC Scopus subject areas

Computer Science(all)
Mathematics(all)

Access to Document

10.1007/978-3-030-39575-9_28Licence: Unspecified

Cite this

Sheikh, T. S., Khan, A., Fahim, M., & Ahmad, M. (2020). Synthesizing data using variational autoencoders for handling class imbalanced deep learning. In W. M. P. van der Aalst, V. Batagelj, D. I. Ignatov, V. Kuskova, S. O. Kuznetsov, I. A. Lomazova, M. Khachay, A. Kutuzov, N. Loukachevitch, A. Napoli, P. M. Pardalos, M. Pelillo, A. V. Savchenko, & E. Tutubalina (Eds.), Analysis of Images, Social Networks and Texts - 8th International Conference, AIST 2019, Revised Selected Papers: Proceedings (pp. 270-281). (Communications in Computer and Information Science; Vol. 1086CCIS). Springer. https://doi.org/10.1007/978-3-030-39575-9_28

Sheikh, Taimoor Shakeel ; Khan, Adil ; Fahim, Muhammad ; Ahmad, Muhammad. / Synthesizing data using variational autoencoders for handling class imbalanced deep learning. Analysis of Images, Social Networks and Texts - 8th International Conference, AIST 2019, Revised Selected Papers: Proceedings. editor / Wil M.P. van der Aalst ; Vladimir Batagelj ; Dmitry I. Ignatov ; Valentina Kuskova ; Sergei O. Kuznetsov ; Irina A. Lomazova ; Michael Khachay ; Andrey Kutuzov ; Natalia Loukachevitch ; Amedeo Napoli ; Panos M. Pardalos ; Marcello Pelillo ; Andrey V. Savchenko ; Elena Tutubalina. Springer, 2020. pp. 270-281 (Communications in Computer and Information Science).

@inproceedings{bdb62df348b24694afad7a3d11837779,

title = "Synthesizing data using variational autoencoders for handling class imbalanced deep learning",

abstract = "This paper addresses the complex problem of learning from unbalanced datasets due to which traditional algorithms may perform poorly. Classification algorithms used for learning tend to favor the larger, less important classes in such problems. In this work, to handle unbalanced data problem, we synthesize data using variational autoencoders (VAE) on raw training samples and then, use various input sources (raw, combination of raw and synthetic) to train different models. We evaluate our method using multiple criteria on SVHN dataset which consists of complex images, and perform a comprehensive comparative analysis of popular CNN architectures when there is balanced and unbalanced data and determine which operates best in class imbalance problem. We found that data synthesis via VAE is reliable and robust, and can help to classify real data with higher accuracy than traditional (unbalanced) data. Our results demonstrate the strength of using VAE to solve the class imbalance problem.",

keywords = "Convolutional Neural Network (CNN), Imbalanced data, Variational autoencoder (VAE)",

author = "Sheikh, {Taimoor Shakeel} and Adil Khan and Muhammad Fahim and Muhammad Ahmad",

note = "Publisher Copyright: {\textcopyright} Springer Nature Switzerland AG 2020. Copyright: Copyright 2020 Elsevier B.V., All rights reserved.; 8th International Conference on Analysis of Images, Social Networks and Texts, AIST 2019 ; Conference date: 17-07-2019 Through 19-07-2019",

year = "2020",

month = feb,

day = "2",

doi = "10.1007/978-3-030-39575-9_28",

language = "English",

isbn = "9783030395742",

series = "Communications in Computer and Information Science",

publisher = "Springer",

pages = "270--281",

editor = "{van der Aalst}, {Wil M.P.} and Vladimir Batagelj and Ignatov, {Dmitry I.} and Valentina Kuskova and Kuznetsov, {Sergei O.} and Lomazova, {Irina A.} and Michael Khachay and Andrey Kutuzov and Natalia Loukachevitch and Amedeo Napoli and Pardalos, {Panos M.} and Marcello Pelillo and Savchenko, {Andrey V.} and Elena Tutubalina",

booktitle = "Analysis of Images, Social Networks and Texts - 8th International Conference, AIST 2019, Revised Selected Papers: Proceedings",

}

Sheikh, TS, Khan, A, Fahim, M & Ahmad, M 2020, Synthesizing data using variational autoencoders for handling class imbalanced deep learning. in WMP van der Aalst, V Batagelj, DI Ignatov, V Kuskova, SO Kuznetsov, IA Lomazova, M Khachay, A Kutuzov, N Loukachevitch, A Napoli, PM Pardalos, M Pelillo, AV Savchenko & E Tutubalina (eds), Analysis of Images, Social Networks and Texts - 8th International Conference, AIST 2019, Revised Selected Papers: Proceedings. Communications in Computer and Information Science, vol. 1086CCIS, Springer, pp. 270-281, 8th International Conference on Analysis of Images, Social Networks and Texts, AIST 2019, Kazan, Russian Federation, 17/07/2019. https://doi.org/10.1007/978-3-030-39575-9_28

Synthesizing data using variational autoencoders for handling class imbalanced deep learning. / Sheikh, Taimoor Shakeel; Khan, Adil; Fahim, Muhammad; Ahmad, Muhammad.

Analysis of Images, Social Networks and Texts - 8th International Conference, AIST 2019, Revised Selected Papers: Proceedings. ed. / Wil M.P. van der Aalst; Vladimir Batagelj; Dmitry I. Ignatov; Valentina Kuskova; Sergei O. Kuznetsov; Irina A. Lomazova; Michael Khachay; Andrey Kutuzov; Natalia Loukachevitch; Amedeo Napoli; Panos M. Pardalos; Marcello Pelillo; Andrey V. Savchenko; Elena Tutubalina. Springer, 2020. p. 270-281 (Communications in Computer and Information Science; Vol. 1086CCIS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Synthesizing data using variational autoencoders for handling class imbalanced deep learning

AU - Sheikh, Taimoor Shakeel

AU - Khan, Adil

AU - Fahim, Muhammad

AU - Ahmad, Muhammad

PY - 2020/2/2

Y1 - 2020/2/2

N2 - This paper addresses the complex problem of learning from unbalanced datasets due to which traditional algorithms may perform poorly. Classification algorithms used for learning tend to favor the larger, less important classes in such problems. In this work, to handle unbalanced data problem, we synthesize data using variational autoencoders (VAE) on raw training samples and then, use various input sources (raw, combination of raw and synthetic) to train different models. We evaluate our method using multiple criteria on SVHN dataset which consists of complex images, and perform a comprehensive comparative analysis of popular CNN architectures when there is balanced and unbalanced data and determine which operates best in class imbalance problem. We found that data synthesis via VAE is reliable and robust, and can help to classify real data with higher accuracy than traditional (unbalanced) data. Our results demonstrate the strength of using VAE to solve the class imbalance problem.

AB - This paper addresses the complex problem of learning from unbalanced datasets due to which traditional algorithms may perform poorly. Classification algorithms used for learning tend to favor the larger, less important classes in such problems. In this work, to handle unbalanced data problem, we synthesize data using variational autoencoders (VAE) on raw training samples and then, use various input sources (raw, combination of raw and synthetic) to train different models. We evaluate our method using multiple criteria on SVHN dataset which consists of complex images, and perform a comprehensive comparative analysis of popular CNN architectures when there is balanced and unbalanced data and determine which operates best in class imbalance problem. We found that data synthesis via VAE is reliable and robust, and can help to classify real data with higher accuracy than traditional (unbalanced) data. Our results demonstrate the strength of using VAE to solve the class imbalance problem.

KW - Convolutional Neural Network (CNN)

KW - Imbalanced data

KW - Variational autoencoder (VAE)

U2 - 10.1007/978-3-030-39575-9_28

DO - 10.1007/978-3-030-39575-9_28

M3 - Conference contribution

AN - SCOPUS:85090491167

SN - 9783030395742

T3 - Communications in Computer and Information Science

SP - 270

EP - 281

BT - Analysis of Images, Social Networks and Texts - 8th International Conference, AIST 2019, Revised Selected Papers: Proceedings

A2 - van der Aalst, Wil M.P.

A2 - Batagelj, Vladimir

A2 - Ignatov, Dmitry I.

A2 - Kuskova, Valentina

A2 - Kuznetsov, Sergei O.

A2 - Lomazova, Irina A.

A2 - Khachay, Michael

A2 - Kutuzov, Andrey

A2 - Loukachevitch, Natalia

A2 - Napoli, Amedeo

A2 - Pardalos, Panos M.

A2 - Pelillo, Marcello

A2 - Savchenko, Andrey V.

A2 - Tutubalina, Elena

PB - Springer

T2 - 8th International Conference on Analysis of Images, Social Networks and Texts, AIST 2019

Y2 - 17 July 2019 through 19 July 2019

ER -

Sheikh TS, Khan A, Fahim M, Ahmad M. Synthesizing data using variational autoencoders for handling class imbalanced deep learning. In van der Aalst WMP, Batagelj V, Ignatov DI, Kuskova V, Kuznetsov SO, Lomazova IA, Khachay M, Kutuzov A, Loukachevitch N, Napoli A, Pardalos PM, Pelillo M, Savchenko AV, Tutubalina E, editors, Analysis of Images, Social Networks and Texts - 8th International Conference, AIST 2019, Revised Selected Papers: Proceedings. Springer. 2020. p. 270-281. (Communications in Computer and Information Science). https://doi.org/10.1007/978-3-030-39575-9_28

Synthesizing data using variational autoencoders for handling class imbalanced deep learning

Abstract

Publication series

Conference

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Fingerprint

Cite this