DecisioNet: A Binary-Tree Structured Neural Network

Noam Gottlieb; Michael Werman

doi:10.1007/978-3-031-26319-4_33

DecisioNet: A Binary-Tree Structured Neural Network

Noam Gottlieb^*, Michael Werman

^*Corresponding author for this work

The Rachel and Selim Benin School of Engineering and Computer Science

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

Abstract

Deep neural networks (DNNs) and decision trees (DTs) are both state-of-the-art classifiers. DNNs perform well due to their representational learning capabilities, while DTs are computationally efficient as they perform inference along one route (root-to-leaf) that is dependent on the input data. In this paper, we present DecisioNet (DN), a binary-tree structured neural network. We propose a systematic way to convert an existing DNN into a DN to create a lightweight version of the original model. DecisioNet takes the best of both worlds - it uses neural modules to perform representational learning and utilizes its tree structure to perform only a portion of the computations. We evaluate various DN architectures, along with their corresponding baseline models on the FashionMNIST, CIFAR10, and CIFAR100 datasets. We show that the DN variants achieve similar accuracy while significantly reducing the computational cost of the original network.

Original language	English
Title of host publication	Computer Vision – ACCV 2022 - 16th Asian Conference on Computer Vision, 2022, Proceedings
Editors	Lei Wang, Juergen Gall, Tat-Jun Chin, Imari Sato, Rama Chellappa
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	556-570
Number of pages	15
ISBN (Print)	9783031263187
DOIs	https://doi.org/10.1007/978-3-031-26319-4_33
State	Published - 2023
Event	16th Asian Conference on Computer Vision, ACCV 2022 - Macao, China Duration: 4 Dec 2022 → 8 Dec 2022

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13841 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	16th Asian Conference on Computer Vision, ACCV 2022
Country/Territory	China
City	Macao
Period	4/12/22 → 8/12/22

Bibliographical note

Funding Information:
Thanks to the ISF (1439/22) and the DFG for funding.

Publisher Copyright:
© 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Keywords

Decision trees
Neural network optimization

Access to Document

10.1007/978-3-031-26319-4_33

Cite this

Gottlieb, N., & Werman, M. (2023). DecisioNet: A Binary-Tree Structured Neural Network. In L. Wang, J. Gall, T.-J. Chin, I. Sato, & R. Chellappa (Eds.), Computer Vision – ACCV 2022 - 16th Asian Conference on Computer Vision, 2022, Proceedings (pp. 556-570). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13841 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-26319-4_33

Gottlieb, Noam ; Werman, Michael. / DecisioNet : A Binary-Tree Structured Neural Network. Computer Vision – ACCV 2022 - 16th Asian Conference on Computer Vision, 2022, Proceedings. editor / Lei Wang ; Juergen Gall ; Tat-Jun Chin ; Imari Sato ; Rama Chellappa. Springer Science and Business Media Deutschland GmbH, 2023. pp. 556-570 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{4fe40737c74b4a128f8adf952071e5b4,

title = "DecisioNet: A Binary-Tree Structured Neural Network",

abstract = "Deep neural networks (DNNs) and decision trees (DTs) are both state-of-the-art classifiers. DNNs perform well due to their representational learning capabilities, while DTs are computationally efficient as they perform inference along one route (root-to-leaf) that is dependent on the input data. In this paper, we present DecisioNet (DN), a binary-tree structured neural network. We propose a systematic way to convert an existing DNN into a DN to create a lightweight version of the original model. DecisioNet takes the best of both worlds - it uses neural modules to perform representational learning and utilizes its tree structure to perform only a portion of the computations. We evaluate various DN architectures, along with their corresponding baseline models on the FashionMNIST, CIFAR10, and CIFAR100 datasets. We show that the DN variants achieve similar accuracy while significantly reducing the computational cost of the original network.",

keywords = "Decision trees, Neural network optimization",

author = "Noam Gottlieb and Michael Werman",

note = "Publisher Copyright: {\textcopyright} 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 16th Asian Conference on Computer Vision, ACCV 2022 ; Conference date: 04-12-2022 Through 08-12-2022",

year = "2023",

doi = "10.1007/978-3-031-26319-4_33",

language = "אנגלית",

isbn = "9783031263187",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "556--570",

editor = "Lei Wang and Juergen Gall and Tat-Jun Chin and Imari Sato and Rama Chellappa",

booktitle = "Computer Vision – ACCV 2022 - 16th Asian Conference on Computer Vision, 2022, Proceedings",

address = "גרמניה",

}

Gottlieb, N & Werman, M 2023, DecisioNet: A Binary-Tree Structured Neural Network. in L Wang, J Gall, T-J Chin, I Sato & R Chellappa (eds), Computer Vision – ACCV 2022 - 16th Asian Conference on Computer Vision, 2022, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13841 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 556-570, 16th Asian Conference on Computer Vision, ACCV 2022, Macao, China, 4/12/22. https://doi.org/10.1007/978-3-031-26319-4_33

DecisioNet: A Binary-Tree Structured Neural Network. / Gottlieb, Noam; Werman, Michael.
Computer Vision – ACCV 2022 - 16th Asian Conference on Computer Vision, 2022, Proceedings. ed. / Lei Wang; Juergen Gall; Tat-Jun Chin; Imari Sato; Rama Chellappa. Springer Science and Business Media Deutschland GmbH, 2023. p. 556-570 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13841 LNCS).

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

TY - GEN

T1 - DecisioNet

T2 - 16th Asian Conference on Computer Vision, ACCV 2022

AU - Gottlieb, Noam

AU - Werman, Michael

PY - 2023

Y1 - 2023

N2 - Deep neural networks (DNNs) and decision trees (DTs) are both state-of-the-art classifiers. DNNs perform well due to their representational learning capabilities, while DTs are computationally efficient as they perform inference along one route (root-to-leaf) that is dependent on the input data. In this paper, we present DecisioNet (DN), a binary-tree structured neural network. We propose a systematic way to convert an existing DNN into a DN to create a lightweight version of the original model. DecisioNet takes the best of both worlds - it uses neural modules to perform representational learning and utilizes its tree structure to perform only a portion of the computations. We evaluate various DN architectures, along with their corresponding baseline models on the FashionMNIST, CIFAR10, and CIFAR100 datasets. We show that the DN variants achieve similar accuracy while significantly reducing the computational cost of the original network.

AB - Deep neural networks (DNNs) and decision trees (DTs) are both state-of-the-art classifiers. DNNs perform well due to their representational learning capabilities, while DTs are computationally efficient as they perform inference along one route (root-to-leaf) that is dependent on the input data. In this paper, we present DecisioNet (DN), a binary-tree structured neural network. We propose a systematic way to convert an existing DNN into a DN to create a lightweight version of the original model. DecisioNet takes the best of both worlds - it uses neural modules to perform representational learning and utilizes its tree structure to perform only a portion of the computations. We evaluate various DN architectures, along with their corresponding baseline models on the FashionMNIST, CIFAR10, and CIFAR100 datasets. We show that the DN variants achieve similar accuracy while significantly reducing the computational cost of the original network.

KW - Decision trees

KW - Neural network optimization

UR - http://www.scopus.com/inward/record.url?scp=85151050970&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-26319-4_33

DO - 10.1007/978-3-031-26319-4_33

M3 - ???researchoutput.researchoutputtypes.contributiontobookanthology.conference???

AN - SCOPUS:85151050970

SN - 9783031263187

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 556

EP - 570

BT - Computer Vision – ACCV 2022 - 16th Asian Conference on Computer Vision, 2022, Proceedings

A2 - Wang, Lei

A2 - Gall, Juergen

A2 - Chin, Tat-Jun

A2 - Sato, Imari

A2 - Chellappa, Rama

PB - Springer Science and Business Media Deutschland GmbH

Y2 - 4 December 2022 through 8 December 2022

ER -

Gottlieb N, Werman M. DecisioNet: A Binary-Tree Structured Neural Network. In Wang L, Gall J, Chin TJ, Sato I, Chellappa R, editors, Computer Vision – ACCV 2022 - 16th Asian Conference on Computer Vision, 2022, Proceedings. Springer Science and Business Media Deutschland GmbH. 2023. p. 556-570. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-26319-4_33

DecisioNet: A Binary-Tree Structured Neural Network

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this