Generalisation performance vs. architecture variations in constructive cascade networks

Suisin Khoo; Tom Gedeon

doi:10.1007/978-3-642-03040-6_29

Generalisation performance vs. architecture variations in constructive cascade networks

Suisin Khoo^*, Tom Gedeon

^*Corresponding author for this work

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

1 Citation (Scopus)

Abstract

Constructive cascade algorithms are powerful methods for training feedforward neural networks with automation of the task of specifying the size and topology of network to use. A series of empirical studies were performed to examine the effect of imposing constraints on constructive cascade neural network architectures. Building a priori knowledge of the task into the network gives better generalisation performance. We introduce our Local Feature Constructive Cascade (LoCC) and Symmetry Local Feature Constructive Cascade (SymLoCC) algorithms, and show them to have good generalisation and network construction properties on face recognition tasks.

Original language	English
Title of host publication	Advances in Neuro-Information Processing - 15th International Conference, ICONIP 2008, Revised Selected Papers
Pages	236-243
Number of pages	8
Edition	PART 2
DOIs	https://doi.org/10.1007/978-3-642-03040-6_29
Publication status	Published - 2009
Event	15th International Conference on Neuro-Information Processing, ICONIP 2008 - Auckland, New Zealand Duration: 25 Nov 2008 → 28 Nov 2008

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Number	PART 2
Volume	5507 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	15th International Conference on Neuro-Information Processing, ICONIP 2008
Country/Territory	New Zealand
City	Auckland
Period	25/11/08 → 28/11/08

Access to Document

10.1007/978-3-642-03040-6_29

Cite this

Khoo, S., & Gedeon, T. (2009). Generalisation performance vs. architecture variations in constructive cascade networks. In Advances in Neuro-Information Processing - 15th International Conference, ICONIP 2008, Revised Selected Papers (PART 2 ed., pp. 236-243). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5507 LNCS, No. PART 2). https://doi.org/10.1007/978-3-642-03040-6_29

Khoo, Suisin ; Gedeon, Tom. / Generalisation performance vs. architecture variations in constructive cascade networks. Advances in Neuro-Information Processing - 15th International Conference, ICONIP 2008, Revised Selected Papers. PART 2. ed. 2009. pp. 236-243 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 2).

@inproceedings{2a6a3133276043ceb81e001f5e0a5fcb,

title = "Generalisation performance vs. architecture variations in constructive cascade networks",

abstract = "Constructive cascade algorithms are powerful methods for training feedforward neural networks with automation of the task of specifying the size and topology of network to use. A series of empirical studies were performed to examine the effect of imposing constraints on constructive cascade neural network architectures. Building a priori knowledge of the task into the network gives better generalisation performance. We introduce our Local Feature Constructive Cascade (LoCC) and Symmetry Local Feature Constructive Cascade (SymLoCC) algorithms, and show them to have good generalisation and network construction properties on face recognition tasks.",

author = "Suisin Khoo and Tom Gedeon",

year = "2009",

doi = "10.1007/978-3-642-03040-6_29",

language = "English",

isbn = "3642030394",

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

number = "PART 2",

pages = "236--243",

booktitle = "Advances in Neuro-Information Processing - 15th International Conference, ICONIP 2008, Revised Selected Papers",

edition = "PART 2",

note = "15th International Conference on Neuro-Information Processing, ICONIP 2008 ; Conference date: 25-11-2008 Through 28-11-2008",

}

Khoo, S & Gedeon, T 2009, Generalisation performance vs. architecture variations in constructive cascade networks. in Advances in Neuro-Information Processing - 15th International Conference, ICONIP 2008, Revised Selected Papers. PART 2 edn, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 2, vol. 5507 LNCS, pp. 236-243, 15th International Conference on Neuro-Information Processing, ICONIP 2008, Auckland, New Zealand, 25/11/08. https://doi.org/10.1007/978-3-642-03040-6_29

Generalisation performance vs. architecture variations in constructive cascade networks. / Khoo, Suisin; Gedeon, Tom.
Advances in Neuro-Information Processing - 15th International Conference, ICONIP 2008, Revised Selected Papers. PART 2. ed. 2009. p. 236-243 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5507 LNCS, No. PART 2).

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

TY - GEN

T1 - Generalisation performance vs. architecture variations in constructive cascade networks

AU - Khoo, Suisin

AU - Gedeon, Tom

PY - 2009

Y1 - 2009

N2 - Constructive cascade algorithms are powerful methods for training feedforward neural networks with automation of the task of specifying the size and topology of network to use. A series of empirical studies were performed to examine the effect of imposing constraints on constructive cascade neural network architectures. Building a priori knowledge of the task into the network gives better generalisation performance. We introduce our Local Feature Constructive Cascade (LoCC) and Symmetry Local Feature Constructive Cascade (SymLoCC) algorithms, and show them to have good generalisation and network construction properties on face recognition tasks.

AB - Constructive cascade algorithms are powerful methods for training feedforward neural networks with automation of the task of specifying the size and topology of network to use. A series of empirical studies were performed to examine the effect of imposing constraints on constructive cascade neural network architectures. Building a priori knowledge of the task into the network gives better generalisation performance. We introduce our Local Feature Constructive Cascade (LoCC) and Symmetry Local Feature Constructive Cascade (SymLoCC) algorithms, and show them to have good generalisation and network construction properties on face recognition tasks.

UR - http://www.scopus.com/inward/record.url?scp=70349148220&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-03040-6_29

DO - 10.1007/978-3-642-03040-6_29

M3 - Conference contribution

SN - 3642030394

SN - 9783642030390

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

SP - 236

EP - 243

BT - Advances in Neuro-Information Processing - 15th International Conference, ICONIP 2008, Revised Selected Papers

T2 - 15th International Conference on Neuro-Information Processing, ICONIP 2008

Y2 - 25 November 2008 through 28 November 2008

ER -

Khoo S, Gedeon T. Generalisation performance vs. architecture variations in constructive cascade networks. In Advances in Neuro-Information Processing - 15th International Conference, ICONIP 2008, Revised Selected Papers. PART 2 ed. 2009. p. 236-243. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 2). doi: 10.1007/978-3-642-03040-6_29

Generalisation performance vs. architecture variations in constructive cascade networks

Abstract

Publication series

Conference

Access to Document

Other files and links

Fingerprint

Cite this