High speed k-winner-take-all competitive learning in reconfigurable hardware

Hui Ya Li; Yao Jung Yeh; Wen-Jyi Hwang; Cheng Tsun Yang

doi:10.1007/978-3-642-02568-6_60

High speed k-winner-take-all competitive learning in reconfigurable hardware

Hui Ya Li, Yao Jung Yeh, Wen-Jyi Hwang, Cheng Tsun Yang

Department of Computer science & Information Engineering

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

4 Citations (Scopus)

Abstract

A novel hardware architecture of the competitive learning (CL) algorithm with k-winners-take-all activation is presented in this paper. It is used as a custom logic block in the arithmetic logic unit (ALU) of the softcore NIOS processor for CL training. Both the partial distance search (PDS) module and hardware divider adopt finite precision calculation for area cost reduction at the expense of slight degradation in training performance. The PDS module also employs subspace search and multiple-coefficient accumulation techniques for effective reduction of the computation latency for the PDS search. Experiment results show that the CPU time is lower than that of Pentium IV processors running the CL training program without the support of custom hardware.

Original language	English
Title of host publication	Next-Generation Applied Intelligence - 22nd International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems, IEA/AIE 2009, Proceedings
Pages	594-603
Number of pages	10
DOIs	https://doi.org/10.1007/978-3-642-02568-6_60
Publication status	Published - 2009 Nov 16
Event	22nd International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems, IEA/AIE 2009 - Tainan, Taiwan Duration: 2009 Jun 24 → 2009 Jun 27

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	5579 LNAI
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	22nd International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems, IEA/AIE 2009
Country	Taiwan
City	Tainan
Period	09/6/24 → 09/6/27

ASJC Scopus subject areas

Theoretical Computer Science
Computer Science(all)

Access to Document

10.1007/978-3-642-02568-6_60

Fingerprint Dive into the research topics of 'High speed k-winner-take-all competitive learning in reconfigurable hardware'. Together they form a unique fingerprint.

Cite this

Li, H. Y., Yeh, Y. J., Hwang, W-J., & Yang, C. T. (2009). High speed k-winner-take-all competitive learning in reconfigurable hardware. In Next-Generation Applied Intelligence - 22nd International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems, IEA/AIE 2009, Proceedings (pp. 594-603). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5579 LNAI). https://doi.org/10.1007/978-3-642-02568-6_60

High speed k-winner-take-all competitive learning in reconfigurable hardware. / Li, Hui Ya; Yeh, Yao Jung; Hwang, Wen-Jyi; Yang, Cheng Tsun.

Next-Generation Applied Intelligence - 22nd International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems, IEA/AIE 2009, Proceedings. 2009. p. 594-603 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5579 LNAI).

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

Li, HY, Yeh, YJ, Hwang, W-J & Yang, CT 2009, High speed k-winner-take-all competitive learning in reconfigurable hardware. in Next-Generation Applied Intelligence - 22nd International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems, IEA/AIE 2009, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 5579 LNAI, pp. 594-603, 22nd International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems, IEA/AIE 2009, Tainan, Taiwan, 09/6/24. https://doi.org/10.1007/978-3-642-02568-6_60

Li HY, Yeh YJ, Hwang W-J, Yang CT. High speed k-winner-take-all competitive learning in reconfigurable hardware. In Next-Generation Applied Intelligence - 22nd International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems, IEA/AIE 2009, Proceedings. 2009. p. 594-603. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-642-02568-6_60

Li, Hui Ya ; Yeh, Yao Jung ; Hwang, Wen-Jyi ; Yang, Cheng Tsun. / High speed k-winner-take-all competitive learning in reconfigurable hardware. Next-Generation Applied Intelligence - 22nd International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems, IEA/AIE 2009, Proceedings. 2009. pp. 594-603 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{c40299932bd847c683fb580fe1199424,

title = "High speed k-winner-take-all competitive learning in reconfigurable hardware",

abstract = "A novel hardware architecture of the competitive learning (CL) algorithm with k-winners-take-all activation is presented in this paper. It is used as a custom logic block in the arithmetic logic unit (ALU) of the softcore NIOS processor for CL training. Both the partial distance search (PDS) module and hardware divider adopt finite precision calculation for area cost reduction at the expense of slight degradation in training performance. The PDS module also employs subspace search and multiple-coefficient accumulation techniques for effective reduction of the computation latency for the PDS search. Experiment results show that the CPU time is lower than that of Pentium IV processors running the CL training program without the support of custom hardware.",

author = "Li, {Hui Ya} and Yeh, {Yao Jung} and Wen-Jyi Hwang and Yang, {Cheng Tsun}",

year = "2009",

month = nov,

day = "16",

doi = "10.1007/978-3-642-02568-6_60",

language = "English",

isbn = "3642025676",

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

pages = "594--603",

booktitle = "Next-Generation Applied Intelligence - 22nd International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems, IEA/AIE 2009, Proceedings",

note = "22nd International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems, IEA/AIE 2009 ; Conference date: 24-06-2009 Through 27-06-2009",

}

TY - GEN

T1 - High speed k-winner-take-all competitive learning in reconfigurable hardware

AU - Li, Hui Ya

AU - Yeh, Yao Jung

AU - Hwang, Wen-Jyi

AU - Yang, Cheng Tsun

PY - 2009/11/16

Y1 - 2009/11/16

N2 - A novel hardware architecture of the competitive learning (CL) algorithm with k-winners-take-all activation is presented in this paper. It is used as a custom logic block in the arithmetic logic unit (ALU) of the softcore NIOS processor for CL training. Both the partial distance search (PDS) module and hardware divider adopt finite precision calculation for area cost reduction at the expense of slight degradation in training performance. The PDS module also employs subspace search and multiple-coefficient accumulation techniques for effective reduction of the computation latency for the PDS search. Experiment results show that the CPU time is lower than that of Pentium IV processors running the CL training program without the support of custom hardware.

AB - A novel hardware architecture of the competitive learning (CL) algorithm with k-winners-take-all activation is presented in this paper. It is used as a custom logic block in the arithmetic logic unit (ALU) of the softcore NIOS processor for CL training. Both the partial distance search (PDS) module and hardware divider adopt finite precision calculation for area cost reduction at the expense of slight degradation in training performance. The PDS module also employs subspace search and multiple-coefficient accumulation techniques for effective reduction of the computation latency for the PDS search. Experiment results show that the CPU time is lower than that of Pentium IV processors running the CL training program without the support of custom hardware.

UR - http://www.scopus.com/inward/record.url?scp=71049188555&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=71049188555&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-02568-6_60

DO - 10.1007/978-3-642-02568-6_60

M3 - Conference contribution

AN - SCOPUS:71049188555

SN - 3642025676

SN - 9783642025679

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

SP - 594

EP - 603

BT - Next-Generation Applied Intelligence - 22nd International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems, IEA/AIE 2009, Proceedings

T2 - 22nd International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems, IEA/AIE 2009

Y2 - 24 June 2009 through 27 June 2009

ER -