On a Constant-Time, Low-Complexity Winner-Take-All Neural Network

Yuen Hsien Tseng; Ja Ling Wu

doi:10.1109/12.376175

On a Constant-Time, Low-Complexity Winner-Take-All Neural Network

Yuen Hsien Tseng
, Ja Ling Wu

Research output: Contribution to journal › Article › peer-review

12 Citations (Scopus)

Abstract

A nearly cost-optimal winner-take-all (WTA) neural network derived from a constant-time sorting network is presented. The resultant WTA network has connection complexity [formula omitted] where s is the depth of cascaded sorting networks. Application of the WTA network to other problems such as nonbinary majority is also included.

Original language	English
Pages (from-to)	601-604
Number of pages	4
Journal	IEEE Transactions on Computers
Volume	44
Issue number	4
DOIs	https://doi.org/10.1109/12.376175
Publication status	Published - 1995 Apr
Externally published	Yes

Keywords

Quadratic perceptron
complexity
nonbinary majority
sorting
winner-take-all

ASJC Scopus subject areas

Software
Theoretical Computer Science
Hardware and Architecture
Computational Theory and Mathematics

Access to Document

10.1109/12.376175

Cite this

@article{8de9865e75f141538491c2d7b7ce7df0,

title = "On a Constant-Time, Low-Complexity Winner-Take-All Neural Network",

abstract = "A nearly cost-optimal winner-take-all (WTA) neural network derived from a constant-time sorting network is presented. The resultant WTA network has connection complexity [formula omitted] where s is the depth of cascaded sorting networks. Application of the WTA network to other problems such as nonbinary majority is also included.",

keywords = "Quadratic perceptron, complexity, nonbinary majority, sorting, winner-take-all",

author = "Tseng, \{Yuen Hsien\} and Wu, \{Ja Ling\}",

note = "Funding Information: We wish to thank Prof. Sartaj Sahni for suggesting this problem and the anonymous referee who suggested an alternate way of merging local segments (Section IV.C), that resulted in an improvement of a O(log”*n) factor. The research of Ramakrishna Thurimella was supported in part by National Science Foundation grant CCR-",

year = "1995",

month = apr,

doi = "10.1109/12.376175",

language = "English",

volume = "44",

pages = "601--604",

journal = "IEEE Transactions on Computers",

issn = "0018-9340",

publisher = "IEEE Computer Society",

number = "4",

}

TY - JOUR

T1 - On a Constant-Time, Low-Complexity Winner-Take-All Neural Network

AU - Tseng, Yuen Hsien

AU - Wu, Ja Ling

N1 - Funding Information: We wish to thank Prof. Sartaj Sahni for suggesting this problem and the anonymous referee who suggested an alternate way of merging local segments (Section IV.C), that resulted in an improvement of a O(log”*n) factor. The research of Ramakrishna Thurimella was supported in part by National Science Foundation grant CCR-

PY - 1995/4

Y1 - 1995/4

N2 - A nearly cost-optimal winner-take-all (WTA) neural network derived from a constant-time sorting network is presented. The resultant WTA network has connection complexity [formula omitted] where s is the depth of cascaded sorting networks. Application of the WTA network to other problems such as nonbinary majority is also included.

AB - A nearly cost-optimal winner-take-all (WTA) neural network derived from a constant-time sorting network is presented. The resultant WTA network has connection complexity [formula omitted] where s is the depth of cascaded sorting networks. Application of the WTA network to other problems such as nonbinary majority is also included.

KW - Quadratic perceptron

KW - complexity

KW - nonbinary majority

KW - sorting

KW - winner-take-all

UR - https://www.scopus.com/pages/publications/0029291227

UR - https://www.scopus.com/pages/publications/0029291227#tab=citedBy

U2 - 10.1109/12.376175

DO - 10.1109/12.376175

M3 - Article

AN - SCOPUS:0029291227

SN - 0018-9340

VL - 44

SP - 601

EP - 604

JO - IEEE Transactions on Computers

JF - IEEE Transactions on Computers

IS - 4

ER -

On a Constant-Time, Low-Complexity Winner-Take-All Neural Network

Abstract

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this