Efficiency optimization of trainable feature extractors for a consumer platform

M.C.J. Peemen; B. Mesman; H. Corporaal

doi:10.1007/978-3-642-23687-7_27

Efficiency optimization of trainable feature extractors for a consumer platform

M.C.J. Peemen, B. Mesman, H. Corporaal

Electronic Systems

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

13 Citations (Scopus)

1 Downloads (Pure)

Abstract

This paper proposes an algorithmic optimization for the feature extractors of biologically inspired Convolutional Neural Networks (CNNs). CNNs are successfully used for different visual pattern recognition applications such as OCR, face detection and object classification. These applications require complex networks exceeding 100,000 interconnected computational nodes. To reduce the computational complexity a modified algorithm is proposed; real benchmarks show 65 - 83% reduction, with equal or even better recognition accuracy. Exploiting the available parallelism in CNNs is essential to reduce the computational scaling problems. Therefore the modified version of the algorithm is implemented and evaluated on a GPU platform to demonstrate the suitability on a cost effective parallel platform. A speedup of 2.5x with respect to the standard algorithm is achieved.

Original language	English
Title of host publication	Proceedings of ACIVS'11 : Advanced Concepts for Intelligent Vision systems, 22-25 August 2011, Ghent, Belgium
Place of Publication	Berlin
Publisher	Springer
Pages	293-304
ISBN (Print)	978-3-642-23686-0
DOIs	https://doi.org/10.1007/978-3-642-23687-7_27
Publication status	Published - 2011
Event	13th International Conference on Advanced Concepts for Intelligent Vision Systems (ACIVS 2011) - Ghent, Belgium Duration: 22 Aug 2011 → 25 Aug 2011 Conference number: 13 http://acivs.org/acivs2011

Publication series

Name	Lecture Notes in Computer Science
Volume	6915
ISSN (Print)	0302-9743

Conference

Conference	13th International Conference on Advanced Concepts for Intelligent Vision Systems (ACIVS 2011)
Abbreviated title	ACIVS 2011
Country/Territory	Belgium
City	Ghent
Period	22/08/11 → 25/08/11
Internet address	http://acivs.org/acivs2011

Access to Document

10.1007/978-3-642-23687-7_27

Cite this

@inproceedings{9007a9b45d4d4d11b17d8345d9c32a85,

title = "Efficiency optimization of trainable feature extractors for a consumer platform",

abstract = "This paper proposes an algorithmic optimization for the feature extractors of biologically inspired Convolutional Neural Networks (CNNs). CNNs are successfully used for different visual pattern recognition applications such as OCR, face detection and object classification. These applications require complex networks exceeding 100,000 interconnected computational nodes. To reduce the computational complexity a modified algorithm is proposed; real benchmarks show 65 - 83% reduction, with equal or even better recognition accuracy. Exploiting the available parallelism in CNNs is essential to reduce the computational scaling problems. Therefore the modified version of the algorithm is implemented and evaluated on a GPU platform to demonstrate the suitability on a cost effective parallel platform. A speedup of 2.5x with respect to the standard algorithm is achieved.",

author = "M.C.J. Peemen and B. Mesman and H. Corporaal",

year = "2011",

doi = "10.1007/978-3-642-23687-7_27",

language = "English",

isbn = "978-3-642-23686-0",

series = "Lecture Notes in Computer Science",

publisher = "Springer",

pages = "293--304",

booktitle = "Proceedings of ACIVS'11 : Advanced Concepts for Intelligent Vision systems, 22-25 August 2011, Ghent, Belgium",

address = "Germany",

note = "13th International Conference on Advanced Concepts for Intelligent Vision Systems (ACIVS 2011), ACIVS 2011 ; Conference date: 22-08-2011 Through 25-08-2011",

url = "http://acivs.org/acivs2011",

}

Peemen, MCJ, Mesman, B & Corporaal, H 2011, Efficiency optimization of trainable feature extractors for a consumer platform. in Proceedings of ACIVS'11 : Advanced Concepts for Intelligent Vision systems, 22-25 August 2011, Ghent, Belgium. Lecture Notes in Computer Science, vol. 6915, Springer, Berlin, pp. 293-304, 13th International Conference on Advanced Concepts for Intelligent Vision Systems (ACIVS 2011), Ghent, Belgium, 22/08/11. https://doi.org/10.1007/978-3-642-23687-7_27

Efficiency optimization of trainable feature extractors for a consumer platform. / Peemen, M.C.J.; Mesman, B.; Corporaal, H.
Proceedings of ACIVS'11 : Advanced Concepts for Intelligent Vision systems, 22-25 August 2011, Ghent, Belgium. Berlin: Springer, 2011. p. 293-304 (Lecture Notes in Computer Science; Vol. 6915).

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

TY - GEN

T1 - Efficiency optimization of trainable feature extractors for a consumer platform

AU - Peemen, M.C.J.

AU - Mesman, B.

AU - Corporaal, H.

N1 - Conference code: 13

PY - 2011

Y1 - 2011

N2 - This paper proposes an algorithmic optimization for the feature extractors of biologically inspired Convolutional Neural Networks (CNNs). CNNs are successfully used for different visual pattern recognition applications such as OCR, face detection and object classification. These applications require complex networks exceeding 100,000 interconnected computational nodes. To reduce the computational complexity a modified algorithm is proposed; real benchmarks show 65 - 83% reduction, with equal or even better recognition accuracy. Exploiting the available parallelism in CNNs is essential to reduce the computational scaling problems. Therefore the modified version of the algorithm is implemented and evaluated on a GPU platform to demonstrate the suitability on a cost effective parallel platform. A speedup of 2.5x with respect to the standard algorithm is achieved.

AB - This paper proposes an algorithmic optimization for the feature extractors of biologically inspired Convolutional Neural Networks (CNNs). CNNs are successfully used for different visual pattern recognition applications such as OCR, face detection and object classification. These applications require complex networks exceeding 100,000 interconnected computational nodes. To reduce the computational complexity a modified algorithm is proposed; real benchmarks show 65 - 83% reduction, with equal or even better recognition accuracy. Exploiting the available parallelism in CNNs is essential to reduce the computational scaling problems. Therefore the modified version of the algorithm is implemented and evaluated on a GPU platform to demonstrate the suitability on a cost effective parallel platform. A speedup of 2.5x with respect to the standard algorithm is achieved.

U2 - 10.1007/978-3-642-23687-7_27

DO - 10.1007/978-3-642-23687-7_27

M3 - Conference contribution

SN - 978-3-642-23686-0

T3 - Lecture Notes in Computer Science

SP - 293

EP - 304

BT - Proceedings of ACIVS'11 : Advanced Concepts for Intelligent Vision systems, 22-25 August 2011, Ghent, Belgium

PB - Springer

CY - Berlin

T2 - 13th International Conference on Advanced Concepts for Intelligent Vision Systems (ACIVS 2011)

Y2 - 22 August 2011 through 25 August 2011

ER -

Efficiency optimization of trainable feature extractors for a consumer platform

Abstract

Publication series

Conference

Access to Document

Fingerprint

Cite this