IP Core Implementation of a Self-Organising Neural Network

David Cyril Hendry, A. A. Duncan, N. Lightowler

Research output: Contribution to journalArticle

47 Citations (Scopus)

Abstract

This paper reports on the design issues and subsequent performance of a soft intellectual property (IP) core implementation of a self-organizing neural network. The design is a development of a previous 0.65-mum single silicon chip providing an array of 256 neurons, where each neuron stores a 16 element reference vector. Migrating the design to a soft IP core presents challenges in achieving the required performance as regards area, power, and clock speed. This same migration, however, offers opportunities for parameterizing the design in a manner which permits a single soft core to meet the requirements of many end users. Thus, the number of neurons within the single instruction multiple data (SIMD) array, the number of elements per reference vector, and the number of bits of each such element are defined by synthesis time parameters. The construction of the SIMD array of neurons is presented including performance results as regards power, area, and classifications per second. For typical parameters (256 neurons with 16 elements per reference vector) the design provides over 2000000 classifications per second using a mainstream 0.18-mum digital process. A reduced instruction set computer processor, the Array. controller (AC), provides both the instruction stream And data to the SIMD array of neurons and an interface to a host processor. The design of this processor is discussed with emphasis on the control aspects which permit supply of a continuous instruction stream to the SIMD array and a flexible interface with the host processor.

Original languageEnglish
Pages (from-to)1085-1096
Number of pages11
JournalIEEE Transactions on Neural Networks
Volume14
Issue number5
DOIs
Publication statusPublished - Sep 2003

Keywords

  • self-organizing neural networks
  • very large-scale integration (VLSI)
  • IP core
  • SIMD
  • COMPUTER

Cite this

IP Core Implementation of a Self-Organising Neural Network. / Hendry, David Cyril; Duncan, A. A.; Lightowler, N.

In: IEEE Transactions on Neural Networks, Vol. 14, No. 5, 09.2003, p. 1085-1096.

Research output: Contribution to journalArticle

Hendry, David Cyril ; Duncan, A. A. ; Lightowler, N. / IP Core Implementation of a Self-Organising Neural Network. In: IEEE Transactions on Neural Networks. 2003 ; Vol. 14, No. 5. pp. 1085-1096.
@article{f3148006587b441bb3a80ef78e508824,
title = "IP Core Implementation of a Self-Organising Neural Network",
abstract = "This paper reports on the design issues and subsequent performance of a soft intellectual property (IP) core implementation of a self-organizing neural network. The design is a development of a previous 0.65-mum single silicon chip providing an array of 256 neurons, where each neuron stores a 16 element reference vector. Migrating the design to a soft IP core presents challenges in achieving the required performance as regards area, power, and clock speed. This same migration, however, offers opportunities for parameterizing the design in a manner which permits a single soft core to meet the requirements of many end users. Thus, the number of neurons within the single instruction multiple data (SIMD) array, the number of elements per reference vector, and the number of bits of each such element are defined by synthesis time parameters. The construction of the SIMD array of neurons is presented including performance results as regards power, area, and classifications per second. For typical parameters (256 neurons with 16 elements per reference vector) the design provides over 2000000 classifications per second using a mainstream 0.18-mum digital process. A reduced instruction set computer processor, the Array. controller (AC), provides both the instruction stream And data to the SIMD array of neurons and an interface to a host processor. The design of this processor is discussed with emphasis on the control aspects which permit supply of a continuous instruction stream to the SIMD array and a flexible interface with the host processor.",
keywords = "self-organizing neural networks, very large-scale integration (VLSI), IP core, SIMD, COMPUTER",
author = "Hendry, {David Cyril} and Duncan, {A. A.} and N. Lightowler",
year = "2003",
month = "9",
doi = "10.1109/TNN.2003.816353",
language = "English",
volume = "14",
pages = "1085--1096",
journal = "IEEE Transactions on Neural Networks",
issn = "1045-9227",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "5",

}

TY - JOUR

T1 - IP Core Implementation of a Self-Organising Neural Network

AU - Hendry, David Cyril

AU - Duncan, A. A.

AU - Lightowler, N.

PY - 2003/9

Y1 - 2003/9

N2 - This paper reports on the design issues and subsequent performance of a soft intellectual property (IP) core implementation of a self-organizing neural network. The design is a development of a previous 0.65-mum single silicon chip providing an array of 256 neurons, where each neuron stores a 16 element reference vector. Migrating the design to a soft IP core presents challenges in achieving the required performance as regards area, power, and clock speed. This same migration, however, offers opportunities for parameterizing the design in a manner which permits a single soft core to meet the requirements of many end users. Thus, the number of neurons within the single instruction multiple data (SIMD) array, the number of elements per reference vector, and the number of bits of each such element are defined by synthesis time parameters. The construction of the SIMD array of neurons is presented including performance results as regards power, area, and classifications per second. For typical parameters (256 neurons with 16 elements per reference vector) the design provides over 2000000 classifications per second using a mainstream 0.18-mum digital process. A reduced instruction set computer processor, the Array. controller (AC), provides both the instruction stream And data to the SIMD array of neurons and an interface to a host processor. The design of this processor is discussed with emphasis on the control aspects which permit supply of a continuous instruction stream to the SIMD array and a flexible interface with the host processor.

AB - This paper reports on the design issues and subsequent performance of a soft intellectual property (IP) core implementation of a self-organizing neural network. The design is a development of a previous 0.65-mum single silicon chip providing an array of 256 neurons, where each neuron stores a 16 element reference vector. Migrating the design to a soft IP core presents challenges in achieving the required performance as regards area, power, and clock speed. This same migration, however, offers opportunities for parameterizing the design in a manner which permits a single soft core to meet the requirements of many end users. Thus, the number of neurons within the single instruction multiple data (SIMD) array, the number of elements per reference vector, and the number of bits of each such element are defined by synthesis time parameters. The construction of the SIMD array of neurons is presented including performance results as regards power, area, and classifications per second. For typical parameters (256 neurons with 16 elements per reference vector) the design provides over 2000000 classifications per second using a mainstream 0.18-mum digital process. A reduced instruction set computer processor, the Array. controller (AC), provides both the instruction stream And data to the SIMD array of neurons and an interface to a host processor. The design of this processor is discussed with emphasis on the control aspects which permit supply of a continuous instruction stream to the SIMD array and a flexible interface with the host processor.

KW - self-organizing neural networks

KW - very large-scale integration (VLSI)

KW - IP core

KW - SIMD

KW - COMPUTER

U2 - 10.1109/TNN.2003.816353

DO - 10.1109/TNN.2003.816353

M3 - Article

VL - 14

SP - 1085

EP - 1096

JO - IEEE Transactions on Neural Networks

JF - IEEE Transactions on Neural Networks

SN - 1045-9227

IS - 5

ER -