Accelerated Optimal Topology Search for Two-hidden-layer Neural Networks

Alan Thomas, Simon Walters, Miltiadis Petridis, Mohammad Malekshahi Gheytassi, Robert Morgan

Research output: Chapter in Book/Conference proceeding with ISSN or ISBNConference contribution with ISSN or ISBNResearchpeer-review

Abstract

Two-hidden-layer feedforward neural networks are investigated for the existence of an optimal hidden node ratio. In the experiments, the heuristic n1 =int (0.5nh + 1), where n1 is the number of nodes in the first hidden layer and nh is the total number of hidden nodes, found networks with generalisation errors, on average, just 0.023%-0.056% greater than those found by exhaustive search. This reduced the complexity of an exhaustive search from quadratic, to linear in , with very little penalty. Further reductions in search complexity to logarithmic could be possible using existing methods developed by the Authors
Original languageEnglish
Title of host publicationEngineering Applications of Neural Networks. EANN 2016
Place of PublicationSwitzerland
PublisherSpringer International Publishing
Pages253-266
Number of pages14
ISBN (Electronic)9783319441887
ISBN (Print)9783319441870
DOIs
Publication statusPublished - 19 Aug 2016
EventEngineering Applications of Neural Networks. EANN 2016 - Aberdeen, UK, 2-5 September, 2016
Duration: 19 Aug 2016 → …

Publication series

NameCommunications in Computer and Information Sciences

Conference

ConferenceEngineering Applications of Neural Networks. EANN 2016
Period19/08/16 → …

Fingerprint

Feedforward neural networks
Topology
Neural networks
Experiments

Bibliographical note

The final publication is available at link.springer.com

Keywords

  • Two-hidden-layer feedforward
  • ANN
  • Exhaustive search
  • Optimal topology
  • Optimal node ratio
  • Heurix
  • Universal function approximation

Cite this

Thomas, A., Walters, S., Petridis, M., Malekshahi Gheytassi, M., & Morgan, R. (2016). Accelerated Optimal Topology Search for Two-hidden-layer Neural Networks. In Engineering Applications of Neural Networks. EANN 2016 (pp. 253-266). (Communications in Computer and Information Sciences). Switzerland: Springer International Publishing. https://doi.org/10.1007/978-3-319-44188-7_19
Thomas, Alan ; Walters, Simon ; Petridis, Miltiadis ; Malekshahi Gheytassi, Mohammad ; Morgan, Robert. / Accelerated Optimal Topology Search for Two-hidden-layer Neural Networks. Engineering Applications of Neural Networks. EANN 2016. Switzerland : Springer International Publishing, 2016. pp. 253-266 (Communications in Computer and Information Sciences).
@inproceedings{d3f977a924ac4944bfa55550842462fe,
title = "Accelerated Optimal Topology Search for Two-hidden-layer Neural Networks",
abstract = "Two-hidden-layer feedforward neural networks are investigated for the existence of an optimal hidden node ratio. In the experiments, the heuristic n1 =int (0.5nh + 1), where n1 is the number of nodes in the first hidden layer and nh is the total number of hidden nodes, found networks with generalisation errors, on average, just 0.023{\%}-0.056{\%} greater than those found by exhaustive search. This reduced the complexity of an exhaustive search from quadratic, to linear in , with very little penalty. Further reductions in search complexity to logarithmic could be possible using existing methods developed by the Authors",
keywords = "Two-hidden-layer feedforward, ANN, Exhaustive search, Optimal topology, Optimal node ratio, Heurix, Universal function approximation",
author = "Alan Thomas and Simon Walters and Miltiadis Petridis and {Malekshahi Gheytassi}, Mohammad and Robert Morgan",
note = "The final publication is available at link.springer.com",
year = "2016",
month = "8",
day = "19",
doi = "10.1007/978-3-319-44188-7_19",
language = "English",
isbn = "9783319441870",
series = "Communications in Computer and Information Sciences",
publisher = "Springer International Publishing",
pages = "253--266",
booktitle = "Engineering Applications of Neural Networks. EANN 2016",

}

Thomas, A, Walters, S, Petridis, M, Malekshahi Gheytassi, M & Morgan, R 2016, Accelerated Optimal Topology Search for Two-hidden-layer Neural Networks. in Engineering Applications of Neural Networks. EANN 2016. Communications in Computer and Information Sciences, Springer International Publishing, Switzerland, pp. 253-266, Engineering Applications of Neural Networks. EANN 2016, 19/08/16. https://doi.org/10.1007/978-3-319-44188-7_19

Accelerated Optimal Topology Search for Two-hidden-layer Neural Networks. / Thomas, Alan; Walters, Simon; Petridis, Miltiadis; Malekshahi Gheytassi, Mohammad; Morgan, Robert.

Engineering Applications of Neural Networks. EANN 2016. Switzerland : Springer International Publishing, 2016. p. 253-266 (Communications in Computer and Information Sciences).

Research output: Chapter in Book/Conference proceeding with ISSN or ISBNConference contribution with ISSN or ISBNResearchpeer-review

TY - GEN

T1 - Accelerated Optimal Topology Search for Two-hidden-layer Neural Networks

AU - Thomas, Alan

AU - Walters, Simon

AU - Petridis, Miltiadis

AU - Malekshahi Gheytassi, Mohammad

AU - Morgan, Robert

N1 - The final publication is available at link.springer.com

PY - 2016/8/19

Y1 - 2016/8/19

N2 - Two-hidden-layer feedforward neural networks are investigated for the existence of an optimal hidden node ratio. In the experiments, the heuristic n1 =int (0.5nh + 1), where n1 is the number of nodes in the first hidden layer and nh is the total number of hidden nodes, found networks with generalisation errors, on average, just 0.023%-0.056% greater than those found by exhaustive search. This reduced the complexity of an exhaustive search from quadratic, to linear in , with very little penalty. Further reductions in search complexity to logarithmic could be possible using existing methods developed by the Authors

AB - Two-hidden-layer feedforward neural networks are investigated for the existence of an optimal hidden node ratio. In the experiments, the heuristic n1 =int (0.5nh + 1), where n1 is the number of nodes in the first hidden layer and nh is the total number of hidden nodes, found networks with generalisation errors, on average, just 0.023%-0.056% greater than those found by exhaustive search. This reduced the complexity of an exhaustive search from quadratic, to linear in , with very little penalty. Further reductions in search complexity to logarithmic could be possible using existing methods developed by the Authors

KW - Two-hidden-layer feedforward

KW - ANN

KW - Exhaustive search

KW - Optimal topology

KW - Optimal node ratio

KW - Heurix

KW - Universal function approximation

U2 - 10.1007/978-3-319-44188-7_19

DO - 10.1007/978-3-319-44188-7_19

M3 - Conference contribution with ISSN or ISBN

SN - 9783319441870

T3 - Communications in Computer and Information Sciences

SP - 253

EP - 266

BT - Engineering Applications of Neural Networks. EANN 2016

PB - Springer International Publishing

CY - Switzerland

ER -

Thomas A, Walters S, Petridis M, Malekshahi Gheytassi M, Morgan R. Accelerated Optimal Topology Search for Two-hidden-layer Neural Networks. In Engineering Applications of Neural Networks. EANN 2016. Switzerland: Springer International Publishing. 2016. p. 253-266. (Communications in Computer and Information Sciences). https://doi.org/10.1007/978-3-319-44188-7_19