Simplifying Neural Networks Using Formal Verification

Sumathi Gokulanathan; Alexander Feldsher; Adi Malca; Clark Barrett; Guy Katz

doi:10.1007/978-3-030-55754-6_5

Simplifying Neural Networks Using Formal Verification

Sumathi Gokulanathan, Alexander Feldsher, Adi Malca, Clark Barrett, Guy Katz

The Rachel and Selim Benin School of Engineering and Computer Science

The Hebrew University of Jerusalem

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

Abstract

Deep neural network (DNN) verification is an emerging field, with diverse verification engines quickly becoming available. Demonstrating the effectiveness of these engines on real-world DNNs is an important step towards their wider adoption. We present a tool that can leverage existing verification engines in performing a novel application: neural network simplification, through the reduction of the size of a DNN without harming its accuracy. We report on the work-flow of the simplification process, and demonstrate its potential significance and applicability on a family of real-world DNNs for aircraft collision avoidance, whose sizes we were able to reduce by as much as 10%.

Original language	English
Title of host publication	NASA Formal Methods - 12th International Symposium, NFM 2020, Proceedings
Editors	Ritchie Lee, Susmit Jha, Anastasia Mavridou
Pages	85-93
Number of pages	9
DOIs	https://doi.org/10.1007/978-3-030-55754-6_5
State	Published - 2020
Event	12th International Symposium on NASA Formal Methods, NFM 2020 - Moffett Field, United States Duration: 11 May 2020 → 15 May 2020

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	12229 LNCS

Conference

Conference	12th International Symposium on NASA Formal Methods, NFM 2020
Country/Territory	United States
City	Moffett Field
Period	11/05/20 → 15/05/20

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 3 Good Health and Well-being

Keywords

Deep neural networks
Marabou
Simplification
Verification

All Science Journal Classification (ASJC) codes

Theoretical Computer Science
General Computer Science

Access to Document

10.1007/978-3-030-55754-6_5

Cite this

Gokulanathan, S., Feldsher, A., Malca, A., Barrett, C., & Katz, G. (2020). Simplifying Neural Networks Using Formal Verification. In R. Lee, S. Jha, & A. Mavridou (Eds.), NASA Formal Methods - 12th International Symposium, NFM 2020, Proceedings (pp. 85-93). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12229 LNCS). https://doi.org/10.1007/978-3-030-55754-6_5

Gokulanathan, Sumathi ; Feldsher, Alexander ; Malca, Adi et al. / Simplifying Neural Networks Using Formal Verification. NASA Formal Methods - 12th International Symposium, NFM 2020, Proceedings. editor / Ritchie Lee ; Susmit Jha ; Anastasia Mavridou. 2020. pp. 85-93 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{fe6a738ec10e4fa291ed0ba6c766183d,

title = "Simplifying Neural Networks Using Formal Verification",

abstract = "Deep neural network (DNN) verification is an emerging field, with diverse verification engines quickly becoming available. Demonstrating the effectiveness of these engines on real-world DNNs is an important step towards their wider adoption. We present a tool that can leverage existing verification engines in performing a novel application: neural network simplification, through the reduction of the size of a DNN without harming its accuracy. We report on the work-flow of the simplification process, and demonstrate its potential significance and applicability on a family of real-world DNNs for aircraft collision avoidance, whose sizes we were able to reduce by as much as 10\%.",

keywords = "Deep neural networks, Marabou, Simplification, Verification",

author = "Sumathi Gokulanathan and Alexander Feldsher and Adi Malca and Clark Barrett and Guy Katz",

note = "Publisher Copyright: {\textcopyright} 2020, Springer Nature Switzerland AG.; 12th International Symposium on NASA Formal Methods, NFM 2020 ; Conference date: 11-05-2020 Through 15-05-2020",

year = "2020",

doi = "10.1007/978-3-030-55754-6\_5",

language = "الإنجليزيّة",

isbn = "9783030557539",

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

pages = "85--93",

editor = "Ritchie Lee and Susmit Jha and Anastasia Mavridou",

booktitle = "NASA Formal Methods - 12th International Symposium, NFM 2020, Proceedings",

}

Gokulanathan, S, Feldsher, A, Malca, A, Barrett, C & Katz, G 2020, Simplifying Neural Networks Using Formal Verification. in R Lee, S Jha & A Mavridou (eds), NASA Formal Methods - 12th International Symposium, NFM 2020, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 12229 LNCS, pp. 85-93, 12th International Symposium on NASA Formal Methods, NFM 2020, Moffett Field, United States, 11/05/20. https://doi.org/10.1007/978-3-030-55754-6_5

Simplifying Neural Networks Using Formal Verification. / Gokulanathan, Sumathi; Feldsher, Alexander; Malca, Adi et al.
NASA Formal Methods - 12th International Symposium, NFM 2020, Proceedings. ed. / Ritchie Lee; Susmit Jha; Anastasia Mavridou. 2020. p. 85-93 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12229 LNCS).

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

TY - GEN

T1 - Simplifying Neural Networks Using Formal Verification

AU - Gokulanathan, Sumathi

AU - Feldsher, Alexander

AU - Malca, Adi

AU - Barrett, Clark

AU - Katz, Guy

PY - 2020

Y1 - 2020

N2 - Deep neural network (DNN) verification is an emerging field, with diverse verification engines quickly becoming available. Demonstrating the effectiveness of these engines on real-world DNNs is an important step towards their wider adoption. We present a tool that can leverage existing verification engines in performing a novel application: neural network simplification, through the reduction of the size of a DNN without harming its accuracy. We report on the work-flow of the simplification process, and demonstrate its potential significance and applicability on a family of real-world DNNs for aircraft collision avoidance, whose sizes we were able to reduce by as much as 10%.

AB - Deep neural network (DNN) verification is an emerging field, with diverse verification engines quickly becoming available. Demonstrating the effectiveness of these engines on real-world DNNs is an important step towards their wider adoption. We present a tool that can leverage existing verification engines in performing a novel application: neural network simplification, through the reduction of the size of a DNN without harming its accuracy. We report on the work-flow of the simplification process, and demonstrate its potential significance and applicability on a family of real-world DNNs for aircraft collision avoidance, whose sizes we were able to reduce by as much as 10%.

KW - Deep neural networks

KW - Marabou

KW - Simplification

KW - Verification

UR - http://www.scopus.com/inward/record.url?scp=85089722312&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-55754-6_5

DO - 10.1007/978-3-030-55754-6_5

M3 - منشور من مؤتمر

SN - 9783030557539

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

SP - 85

EP - 93

BT - NASA Formal Methods - 12th International Symposium, NFM 2020, Proceedings

A2 - Lee, Ritchie

A2 - Jha, Susmit

A2 - Mavridou, Anastasia

T2 - 12th International Symposium on NASA Formal Methods, NFM 2020

Y2 - 11 May 2020 through 15 May 2020

ER -

Gokulanathan S, Feldsher A, Malca A, Barrett C, Katz G. Simplifying Neural Networks Using Formal Verification. In Lee R, Jha S, Mavridou A, editors, NASA Formal Methods - 12th International Symposium, NFM 2020, Proceedings. 2020. p. 85-93. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-55754-6_5

Simplifying Neural Networks Using Formal Verification

Abstract

Publication series

Conference

UN SDGs

Keywords

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this