Implementing size-optimal discrete neural networks require analog circuitry

Valeriu Beiu

Implementing size-optimal discrete neural networks require analog circuitry

Valeriu Beiu

Research output: Contribution to journal › Conference article › peer-review

Abstract

This paper starts by overviewing results dealing with the approximation capabilities of neural networks, as well as bounds on the size of threshold gate circuits. Based on a constructive solution for Kolmogorov's superpositions we will show that implementing Boolean functions can be done using neurons having an identity transfer function. Because in this case the size of the network is minimised, it follows that sizeoptimal solutions for implementing Boolean functions can be obtained using analog circuitry. Conclusions and several comments on the required precision are ending the paper.

Original language	English
Journal	European Signal Processing Conference
Volume	1998-January
Publication status	Published - 1998
Externally published	Yes
Event	9th European Signal Processing Conference, EUSIPCO 1998 - Island of Rhodes, Greece Duration: Sept 8 1998 → Sept 11 1998

Keywords

Analog circuits
Kolmogorov's superpositions
Neural networks
Precision
Size
Threshold gate circuits

ASJC Scopus subject areas

Signal Processing
Electrical and Electronic Engineering

Cite this

@article{9b529a72b4d54283ace6c5534e40e955,

title = "Implementing size-optimal discrete neural networks require analog circuitry",

abstract = "This paper starts by overviewing results dealing with the approximation capabilities of neural networks, as well as bounds on the size of threshold gate circuits. Based on a constructive solution for Kolmogorov's superpositions we will show that implementing Boolean functions can be done using neurons having an identity transfer function. Because in this case the size of the network is minimised, it follows that sizeoptimal solutions for implementing Boolean functions can be obtained using analog circuitry. Conclusions and several comments on the required precision are ending the paper.",

keywords = "Analog circuits, Kolmogorov's superpositions, Neural networks, Precision, Size, Threshold gate circuits",

author = "Valeriu Beiu",

year = "1998",

language = "English",

volume = "1998-January",

journal = "European Signal Processing Conference",

issn = "2219-5491",

publisher = "European Signal Processing Conference, EUSIPCO",

note = "9th European Signal Processing Conference, EUSIPCO 1998 ; Conference date: 08-09-1998 Through 11-09-1998",

}

TY - JOUR

T1 - Implementing size-optimal discrete neural networks require analog circuitry

AU - Beiu, Valeriu

PY - 1998

Y1 - 1998

N2 - This paper starts by overviewing results dealing with the approximation capabilities of neural networks, as well as bounds on the size of threshold gate circuits. Based on a constructive solution for Kolmogorov's superpositions we will show that implementing Boolean functions can be done using neurons having an identity transfer function. Because in this case the size of the network is minimised, it follows that sizeoptimal solutions for implementing Boolean functions can be obtained using analog circuitry. Conclusions and several comments on the required precision are ending the paper.

AB - This paper starts by overviewing results dealing with the approximation capabilities of neural networks, as well as bounds on the size of threshold gate circuits. Based on a constructive solution for Kolmogorov's superpositions we will show that implementing Boolean functions can be done using neurons having an identity transfer function. Because in this case the size of the network is minimised, it follows that sizeoptimal solutions for implementing Boolean functions can be obtained using analog circuitry. Conclusions and several comments on the required precision are ending the paper.

KW - Analog circuits

KW - Kolmogorov's superpositions

KW - Neural networks

KW - Precision

KW - Size

KW - Threshold gate circuits

UR - http://www.scopus.com/inward/record.url?scp=85019587588&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85019587588&partnerID=8YFLogxK

M3 - Conference article

AN - SCOPUS:85019587588

SN - 2219-5491

VL - 1998-January

JO - European Signal Processing Conference

JF - European Signal Processing Conference

T2 - 9th European Signal Processing Conference, EUSIPCO 1998

Y2 - 8 September 1998 through 11 September 1998

ER -

Implementing size-optimal discrete neural networks require analog circuitry

Abstract

Keywords

ASJC Scopus subject areas

Other files and links

Fingerprint

Cite this