Centroid-Based Differential Evolution with Composite Trial Vector Generation Strategies for Neural Network Training

Sahar Rahmani; Seyed Jalaleddin Mousavirad; Mohammed El-Abd; Gerald Schaefer; Diego Oliva

doi:10.1007/978-3-031-30229-9_39

Centroid-Based Differential Evolution with Composite Trial Vector Generation Strategies for Neural Network Training

Sahar Rahmani, Seyed Jalaleddin Mousavirad, Mohammed El-Abd, Gerald Schaefer, Diego Oliva

Electrical and Computer Engineering

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

4 Scopus citations

Abstract

The learning process of feedforward neural networks, which determines suitable connection weights and biases, is a challenging machine learning problems and significantly impact how well neural networks work. Back-propagation, a gradient descent-based method, is one of the most popular learning algorithms, but tends to get stuck in local optima. Differential evolution (DE), a popular population-based metaheuristic algorithm, is an interesting alternative for tackling challenging optimisation problems. In this paper, we present Cen-CoDE, a centroid-based differential evolution algorithm with composite trial vector generation strategies and control parameters to train neural networks. Our algorithm encodes weights and biases into a candidate solution, employs a centroid-based strategy in three different ways to generate different trial vectors, while the objective function is based on classification error. In our experiments, we show Cen-CoDE to outperform other contemporary techniques.

Original language	English
Title of host publication	Applications of Evolutionary Computation - 26th European Conference, EvoApplications 2023, Held as Part of EvoStar 2023, Proceedings
Editors	João Correia, Stephen Smith, Raneem Qaddoura
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	608-622
Number of pages	15
ISBN (Print)	9783031302282
DOIs	https://doi.org/10.1007/978-3-031-30229-9_39
State	Published - 2023
Event	26th International Conference on Applications of Evolutionary Computation, EvoApplications 2023, held as part of EvoStar 2023 - Brno, Czech Republic Duration: 12 Apr 2023 → 14 Apr 2023

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13989 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	26th International Conference on Applications of Evolutionary Computation, EvoApplications 2023, held as part of EvoStar 2023
Country/Territory	Czech Republic
City	Brno
Period	12/04/23 → 14/04/23

Keywords

classification
differential evolution
feedforward neural network
learning

Access to Document

10.1007/978-3-031-30229-9_39

Cite this

Rahmani, S., Mousavirad, S. J., El-Abd, M., Schaefer, G., & Oliva, D. (2023). Centroid-Based Differential Evolution with Composite Trial Vector Generation Strategies for Neural Network Training. In J. Correia, S. Smith, & R. Qaddoura (Eds.), Applications of Evolutionary Computation - 26th European Conference, EvoApplications 2023, Held as Part of EvoStar 2023, Proceedings (pp. 608-622). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13989 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-30229-9_39

Rahmani, Sahar ; Mousavirad, Seyed Jalaleddin ; El-Abd, Mohammed et al. / Centroid-Based Differential Evolution with Composite Trial Vector Generation Strategies for Neural Network Training. Applications of Evolutionary Computation - 26th European Conference, EvoApplications 2023, Held as Part of EvoStar 2023, Proceedings. editor / João Correia ; Stephen Smith ; Raneem Qaddoura. Springer Science and Business Media Deutschland GmbH, 2023. pp. 608-622 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{849fdd6c63fb4b7ba355e7a5e5309825,

title = "Centroid-Based Differential Evolution with Composite Trial Vector Generation Strategies for Neural Network Training",

abstract = "The learning process of feedforward neural networks, which determines suitable connection weights and biases, is a challenging machine learning problems and significantly impact how well neural networks work. Back-propagation, a gradient descent-based method, is one of the most popular learning algorithms, but tends to get stuck in local optima. Differential evolution (DE), a popular population-based metaheuristic algorithm, is an interesting alternative for tackling challenging optimisation problems. In this paper, we present Cen-CoDE, a centroid-based differential evolution algorithm with composite trial vector generation strategies and control parameters to train neural networks. Our algorithm encodes weights and biases into a candidate solution, employs a centroid-based strategy in three different ways to generate different trial vectors, while the objective function is based on classification error. In our experiments, we show Cen-CoDE to outperform other contemporary techniques.",

keywords = "classification, differential evolution, feedforward neural network, learning",

author = "Sahar Rahmani and Mousavirad, {Seyed Jalaleddin} and Mohammed El-Abd and Gerald Schaefer and Diego Oliva",

note = "Publisher Copyright: {\textcopyright} 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 26th International Conference on Applications of Evolutionary Computation, EvoApplications 2023, held as part of EvoStar 2023 ; Conference date: 12-04-2023 Through 14-04-2023",

year = "2023",

doi = "10.1007/978-3-031-30229-9_39",

language = "English",

isbn = "9783031302282",

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 = "608--622",

editor = "Jo{\~a}o Correia and Stephen Smith and Raneem Qaddoura",

booktitle = "Applications of Evolutionary Computation - 26th European Conference, EvoApplications 2023, Held as Part of EvoStar 2023, Proceedings",

address = "Germany",

}

Rahmani, S, Mousavirad, SJ, El-Abd, M, Schaefer, G & Oliva, D 2023, Centroid-Based Differential Evolution with Composite Trial Vector Generation Strategies for Neural Network Training. in J Correia, S Smith & R Qaddoura (eds), Applications of Evolutionary Computation - 26th European Conference, EvoApplications 2023, Held as Part of EvoStar 2023, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13989 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 608-622, 26th International Conference on Applications of Evolutionary Computation, EvoApplications 2023, held as part of EvoStar 2023, Brno, Czech Republic, 12/04/23. https://doi.org/10.1007/978-3-031-30229-9_39

Centroid-Based Differential Evolution with Composite Trial Vector Generation Strategies for Neural Network Training. / Rahmani, Sahar; Mousavirad, Seyed Jalaleddin; El-Abd, Mohammed et al.
Applications of Evolutionary Computation - 26th European Conference, EvoApplications 2023, Held as Part of EvoStar 2023, Proceedings. ed. / João Correia; Stephen Smith; Raneem Qaddoura. Springer Science and Business Media Deutschland GmbH, 2023. p. 608-622 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13989 LNCS).

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

TY - GEN

T1 - Centroid-Based Differential Evolution with Composite Trial Vector Generation Strategies for Neural Network Training

AU - Rahmani, Sahar

AU - Mousavirad, Seyed Jalaleddin

AU - El-Abd, Mohammed

AU - Schaefer, Gerald

AU - Oliva, Diego

PY - 2023

Y1 - 2023

N2 - The learning process of feedforward neural networks, which determines suitable connection weights and biases, is a challenging machine learning problems and significantly impact how well neural networks work. Back-propagation, a gradient descent-based method, is one of the most popular learning algorithms, but tends to get stuck in local optima. Differential evolution (DE), a popular population-based metaheuristic algorithm, is an interesting alternative for tackling challenging optimisation problems. In this paper, we present Cen-CoDE, a centroid-based differential evolution algorithm with composite trial vector generation strategies and control parameters to train neural networks. Our algorithm encodes weights and biases into a candidate solution, employs a centroid-based strategy in three different ways to generate different trial vectors, while the objective function is based on classification error. In our experiments, we show Cen-CoDE to outperform other contemporary techniques.

AB - The learning process of feedforward neural networks, which determines suitable connection weights and biases, is a challenging machine learning problems and significantly impact how well neural networks work. Back-propagation, a gradient descent-based method, is one of the most popular learning algorithms, but tends to get stuck in local optima. Differential evolution (DE), a popular population-based metaheuristic algorithm, is an interesting alternative for tackling challenging optimisation problems. In this paper, we present Cen-CoDE, a centroid-based differential evolution algorithm with composite trial vector generation strategies and control parameters to train neural networks. Our algorithm encodes weights and biases into a candidate solution, employs a centroid-based strategy in three different ways to generate different trial vectors, while the objective function is based on classification error. In our experiments, we show Cen-CoDE to outperform other contemporary techniques.

KW - classification

KW - differential evolution

KW - feedforward neural network

KW - learning

UR - http://www.scopus.com/inward/record.url?scp=85159403528&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-30229-9_39

DO - 10.1007/978-3-031-30229-9_39

M3 - Conference contribution

AN - SCOPUS:85159403528

SN - 9783031302282

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

SP - 608

EP - 622

BT - Applications of Evolutionary Computation - 26th European Conference, EvoApplications 2023, Held as Part of EvoStar 2023, Proceedings

A2 - Correia, João

A2 - Smith, Stephen

A2 - Qaddoura, Raneem

PB - Springer Science and Business Media Deutschland GmbH

T2 - 26th International Conference on Applications of Evolutionary Computation, EvoApplications 2023, held as part of EvoStar 2023

Y2 - 12 April 2023 through 14 April 2023

ER -

Rahmani S, Mousavirad SJ, El-Abd M, Schaefer G, Oliva D. Centroid-Based Differential Evolution with Composite Trial Vector Generation Strategies for Neural Network Training. In Correia J, Smith S, Qaddoura R, editors, Applications of Evolutionary Computation - 26th European Conference, EvoApplications 2023, Held as Part of EvoStar 2023, Proceedings. Springer Science and Business Media Deutschland GmbH. 2023. p. 608-622. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-30229-9_39