Evolutionary Optimisation of Fully Connected Artificial Neural Network Topology

Jordan J. Bird*, Anikó Ekárt, Christopher D. Buckingham, Diego R. Faria

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This paper proposes an approach to selecting the amount of layers and neurons contained within Multilayer Perceptron hidden layers through a single-objective evolutionary approach with the goal of model accuracy. At each generation, a population of Neural Network architectures are created and ranked by their accuracy. The generated solutions are combined in a breeding process to create a larger population, and at each generation the weakest solutions are removed to retain the population size inspired by a Darwinian ‘survival of the fittest’. Multiple datasets are tested, and results show that architectures can be successfully improved and derived through a hyper-heuristic evolutionary approach, in less than 10% of the exhaustive search time. The evolutionary approach was further optimised through population density increase as well as gradual solution max complexity increase throughout the simulation.

Original languageEnglish
Title of host publicationIntelligent Computing - Proceedings of the 2019 Computing Conference
EditorsKohei Arai, Rahul Bhatia, Supriya Kapoor
PublisherSpringer-Verlag Wien
Pages751-762
Number of pages12
ISBN (Print)9783030228705
DOIs
Publication statusPublished - 23 Jun 2019
EventComputing Conference, 2019 - London, United Kingdom
Duration: 16 Jul 201917 Jul 2019

Publication series

NameAdvances in Intelligent Systems and Computing
Volume997
ISSN (Print)2194-5357
ISSN (Electronic)2194-5365

Conference

ConferenceComputing Conference, 2019
CountryUnited Kingdom
CityLondon
Period16/07/1917/07/19

Fingerprint

Topology
Neural networks
Multilayer neural networks
Network architecture
Neurons

Keywords

  • Computational intelligence
  • Evolutionary computation
  • Hyperheuristics
  • Neural networks
  • Neuroevolution

Cite this

Bird, J. J., Ekárt, A., Buckingham, C. D., & Faria, D. R. (2019). Evolutionary Optimisation of Fully Connected Artificial Neural Network Topology. In K. Arai, R. Bhatia, & S. Kapoor (Eds.), Intelligent Computing - Proceedings of the 2019 Computing Conference (pp. 751-762). (Advances in Intelligent Systems and Computing; Vol. 997). Springer-Verlag Wien. https://doi.org/10.1007/978-3-030-22871-2_52
Bird, Jordan J. ; Ekárt, Anikó ; Buckingham, Christopher D. ; Faria, Diego R. / Evolutionary Optimisation of Fully Connected Artificial Neural Network Topology. Intelligent Computing - Proceedings of the 2019 Computing Conference. editor / Kohei Arai ; Rahul Bhatia ; Supriya Kapoor. Springer-Verlag Wien, 2019. pp. 751-762 (Advances in Intelligent Systems and Computing).
@inproceedings{4bdeb80d800b4b7f908d6d35763660fc,
title = "Evolutionary Optimisation of Fully Connected Artificial Neural Network Topology",
abstract = "This paper proposes an approach to selecting the amount of layers and neurons contained within Multilayer Perceptron hidden layers through a single-objective evolutionary approach with the goal of model accuracy. At each generation, a population of Neural Network architectures are created and ranked by their accuracy. The generated solutions are combined in a breeding process to create a larger population, and at each generation the weakest solutions are removed to retain the population size inspired by a Darwinian ‘survival of the fittest’. Multiple datasets are tested, and results show that architectures can be successfully improved and derived through a hyper-heuristic evolutionary approach, in less than 10{\%} of the exhaustive search time. The evolutionary approach was further optimised through population density increase as well as gradual solution max complexity increase throughout the simulation.",
keywords = "Computational intelligence, Evolutionary computation, Hyperheuristics, Neural networks, Neuroevolution",
author = "Bird, {Jordan J.} and Anik{\'o} Ek{\'a}rt and Buckingham, {Christopher D.} and Faria, {Diego R.}",
year = "2019",
month = "6",
day = "23",
doi = "10.1007/978-3-030-22871-2_52",
language = "English",
isbn = "9783030228705",
series = "Advances in Intelligent Systems and Computing",
publisher = "Springer-Verlag Wien",
pages = "751--762",
editor = "Kohei Arai and Rahul Bhatia and Supriya Kapoor",
booktitle = "Intelligent Computing - Proceedings of the 2019 Computing Conference",
address = "Austria",

}

Bird, JJ, Ekárt, A, Buckingham, CD & Faria, DR 2019, Evolutionary Optimisation of Fully Connected Artificial Neural Network Topology. in K Arai, R Bhatia & S Kapoor (eds), Intelligent Computing - Proceedings of the 2019 Computing Conference. Advances in Intelligent Systems and Computing, vol. 997, Springer-Verlag Wien, pp. 751-762, Computing Conference, 2019, London, United Kingdom, 16/07/19. https://doi.org/10.1007/978-3-030-22871-2_52

Evolutionary Optimisation of Fully Connected Artificial Neural Network Topology. / Bird, Jordan J.; Ekárt, Anikó; Buckingham, Christopher D.; Faria, Diego R.

Intelligent Computing - Proceedings of the 2019 Computing Conference. ed. / Kohei Arai; Rahul Bhatia; Supriya Kapoor. Springer-Verlag Wien, 2019. p. 751-762 (Advances in Intelligent Systems and Computing; Vol. 997).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Evolutionary Optimisation of Fully Connected Artificial Neural Network Topology

AU - Bird, Jordan J.

AU - Ekárt, Anikó

AU - Buckingham, Christopher D.

AU - Faria, Diego R.

PY - 2019/6/23

Y1 - 2019/6/23

N2 - This paper proposes an approach to selecting the amount of layers and neurons contained within Multilayer Perceptron hidden layers through a single-objective evolutionary approach with the goal of model accuracy. At each generation, a population of Neural Network architectures are created and ranked by their accuracy. The generated solutions are combined in a breeding process to create a larger population, and at each generation the weakest solutions are removed to retain the population size inspired by a Darwinian ‘survival of the fittest’. Multiple datasets are tested, and results show that architectures can be successfully improved and derived through a hyper-heuristic evolutionary approach, in less than 10% of the exhaustive search time. The evolutionary approach was further optimised through population density increase as well as gradual solution max complexity increase throughout the simulation.

AB - This paper proposes an approach to selecting the amount of layers and neurons contained within Multilayer Perceptron hidden layers through a single-objective evolutionary approach with the goal of model accuracy. At each generation, a population of Neural Network architectures are created and ranked by their accuracy. The generated solutions are combined in a breeding process to create a larger population, and at each generation the weakest solutions are removed to retain the population size inspired by a Darwinian ‘survival of the fittest’. Multiple datasets are tested, and results show that architectures can be successfully improved and derived through a hyper-heuristic evolutionary approach, in less than 10% of the exhaustive search time. The evolutionary approach was further optimised through population density increase as well as gradual solution max complexity increase throughout the simulation.

KW - Computational intelligence

KW - Evolutionary computation

KW - Hyperheuristics

KW - Neural networks

KW - Neuroevolution

UR - http://www.scopus.com/inward/record.url?scp=85069202275&partnerID=8YFLogxK

UR - https://link.springer.com/chapter/10.1007%2F978-3-030-22871-2_52

U2 - 10.1007/978-3-030-22871-2_52

DO - 10.1007/978-3-030-22871-2_52

M3 - Conference contribution

AN - SCOPUS:85069202275

SN - 9783030228705

T3 - Advances in Intelligent Systems and Computing

SP - 751

EP - 762

BT - Intelligent Computing - Proceedings of the 2019 Computing Conference

A2 - Arai, Kohei

A2 - Bhatia, Rahul

A2 - Kapoor, Supriya

PB - Springer-Verlag Wien

ER -

Bird JJ, Ekárt A, Buckingham CD, Faria DR. Evolutionary Optimisation of Fully Connected Artificial Neural Network Topology. In Arai K, Bhatia R, Kapoor S, editors, Intelligent Computing - Proceedings of the 2019 Computing Conference. Springer-Verlag Wien. 2019. p. 751-762. (Advances in Intelligent Systems and Computing). https://doi.org/10.1007/978-3-030-22871-2_52