Implementing Kak Neural Networks on a Reconfigurable Computing Platform

Lecture notes in computer sciencePublished 1 January 2000

Jihan Zhu, George Milne

DOI10.1007/3-540-44614-1_29

Source URL

Citations11

SJR quartileQ2

SJR score0.35

SNIP0.55

TL;DR

This paper shows that the Kak algorithm is hardware friendly and is especially suited for implementation in reconfigurable computing using fine grained parallelism and demonstrates that on-line learning with the algorithm is possible through dynamic evolution of the topology of a Kak neural network.

Abstract

The training of neural networks occurs instantaneously with Kak's corner classification algorithm CC4. It is based on prescriptive learning, hence is extremely fast compared with iterative supervised learning algorithms such as backpropagation. This paper shows that the Kak algorithm is hardware friendly and is especially suited for implementation in reconfigurable computing using fine grained parallelism. We also demonstrate that on-line learning with the algorithm is possible through dynamic evolution of the topology of a Kak neural network.

Keywords

Computer ScienceEngineering

IEEE Signal Processing MagazineProgress in supervised neural networks

1,246 Citations1993D. Hush, B.G. Horne

Theoretical results concerning the capabilities and limitations of various neural network models are summarized, and some of their extensions are discussed.

Design methodologies for partially reconfigured systems

97 Citations2002Jason Hadley, Brad Hutchings

The paper presents the design methodology for partial RTR in the context of RRANN2, apartial RTR artificial neural network, and shows how this reduces the amount of time that an RTR application spends configuring and significantly enhances overall performance.

Density enhancement of a neural network using FPGAs and run-time reconfiguration

96 Citations2002J.G. Eldredge, Brad Hutchings

The run-time reconfiguration artificial neural network (RRANN) uses ran-time resurfacing to increase the hardware density of FPGAs and is a flexible realization of the time/space trade-off.

Lecture notes in computer scienceArtificial neural network implementation on a fine-grained FPGA

52 Citations1994Patrick Lysaght, J. Stockwood +2 more

This paper reports on the implementation of an Artificial Neural Network (ANN) on an Atmel AT6005 Field Programmable Gate Array (FPGA) as an experiment in mapping a bit-level, logically intensive application onto the specific logic resources of a fine-grained FPGA.

Information SciencesOn generalization by neural networks

45 Citations1998Subhash Kak

It is shown that a prescriptive learning procedure where the weights are simply read off based on the training data can provide good generalization.

Circuits Systems and Signal ProcessingA new corner classification approach to neural network training

31 Citations1998Kun-Won Tang, Subhash Kak

A new and improved corner classification technique that retains the prescriptive learning capability and gives excellent generalization performance is presented that could be the basis of the recently introduced neuroscientific notion of “working memory.”

Realising massively concurrent systems on the SPACE machine

24 Citations2002George Milne, Paul Cockshott +2 more

A cellular automata type of architecture is introduced which is used to simulate highly concurrent systems directly in hardware, with physical concurrency being realised by true hardware concurrency.

Information SciencesComparison of learning and generalization capabilities of the Kak and the Backpropagation algorithms

23 Citations1994Praveen Raina

It is found that the Backpropagation algorithm is much more computation-intensive than the Kak algorithms that have comparable generalization performance.