Spectral technique for hidden layer neural network training
Pattern Recognition LettersPublished 1 August 1997
Terry Windeatt, Robert Tebbs
Citations13
SJR quartileQ1
SJR score1.00
SNIP1.43
Generate an AI Snapshot to get a quick, structured summary of this paper.
Study Snapshot
ObjectiveStudy objective
MethodsResearch methodology
PopulationPopulation studied
Sample sizeSample sizes
OutcomesStudy outcomes here
ResultsStudy results comes here
LimitationsResearch study limitations comes here
A concise AI-generated summary of the paper will appear here once you click Generate AI Snapshot.
TL;DR
Weight constraints derived from a spectral summation are used to check separability during the partitioning phase, and to limit hyperplane movement during training.
Abstract
We propose a new constructive algorithm for learning binary-to-binary mappings. Weight constraints derived from a spectral summation are used to check separability during the partitioning phase, and to limit hyperplane movement during training.
Keywords
Computer Science
IBM Journal of Research and DevelopmentDiagnosis of Automata Failures: A Calculus and a Method
889 Citations1966J. P. Roth
The problem considered is the diagnosis of failures of automata, specifically, failures that manifest themselves as logical malfunctions, and an algorithm is developed which utilizes this calculus to compute tests to detect failures.
IEEE Transactions on Neural NetworksPerceptron-based learning algorithms
551 Citations1990Stephen I. Gallant
The heart of these algorithms is the pocket algorithm, a modification of perceptron learning that makes perceptronLearning well-behaved with nonseparable training data, even if the data are noisy and contradictory.
Europhysics Letters (EPL)A Convergence Theorem for Sequential Learning in Two-Layer Perceptrons
136 Citations1990Mario Marchand, Mostefa Golea +1 more
It is proved that the internal representations obtained by such procedures are linearly separable and compare favourably with results of other growth algorithms.
Neural ComputationA "Thermal" Perceptron Learning Rule
102 Citations1992Marcus Frean
The thermal perceptron is a simple extension to Rosenblatt's perceptron learning rule for training individual linear threshold units that finds stable weights for nonseparable problems as well as separable ones.
IEEE Transactions on Computer-Aided Design of Integrated Circuits and SystemsEffective computer methods for the calculation of Rademacher-Walsh spectrum for completely and incompletely specified Boolean functions
69 Citations1992B.J. Falkowski, Ingo Schäfer +1 more
The algorithm and its implementation provide the fastest and most comprehensive program (having many options) known to the authors for the calculation of the Rademacher-Walsh transform.
Network Computation in Neural SystemsOn learning simple neural concepts: from halfspace intersections to neural decision lists
45 Citations1993Mario Marchand, Mostefa Golea
It is shown how the greedy method can be generalized to handle this class of concepts, the class of halfspace intersections, which shows that these concepts are not only important from the theoretical point of view, but also in practice.
IEEE Transactions on Neural NetworksA constructive algorithm for binary neural networks: the oil-spot algorithm
41 Citations1995Fabio Massimo Frattale Mascioli, Giovanni Martinelli
A constructive training algorithm for supervised neural networks that dynamically constructs a two-layer neural network by involving successively binary examples based on the representation of the mapping of interest onto the binary hypercube of the input space.
IEEE Transactions on Neural NetworksOn sequential construction of binary neural networks
33 Citations1995Marco Muselli
A new technique called sequential window learning (SWL), for the construction of two-layer perceptrons with binary inputs is presented, and the introduction of a new type of neuron, having a window-shaped activation function, considerably increases the convergence speed and the compactness of resulting networks.
Europhysics Letters (EPL)A Fast Partitioning Algorithm and a Comparison of Binary Feedforward Neural Networks
27 Citations1992S. A. J Keibek, G. T. Barkema +3 more
A comparison was carried out of several learning algorithms for training feedforward neural networks with linear threshold units and a fast method for the selection of input patterns that can be identified by a single neuron is presented.
Network Computation in Neural SystemsOn learning simple neural concepts: from halfspace intersections to neural decision lists
26 Citations1993Mario Marchand, Mostefa Golea
IEEE Design & Test of ComputersTest counting: a tool for VLSI testing
25 Citations1989Sheldon B. Akers, Balakrishnan Krishnamurthy
The authors present a technique, called test counting, for analyzing the testing requirements imposed on a combinational network in the form of a set of stuck-at faults to be detected by solving a large set of mathematical inequalities.
Network Computation in Neural SystemsThe Patch algorithm: fast design of binary feedforward neural networks
13 Citations1993G. T. Barkema, H.M.A. Andree +1 more
A new constructive learning algorithm to generate binary neural networks is presented, able to handle analogue inputs and problems with multiple output states and allows training sets up to several thousands of patterns.
Network Computation in Neural SystemsThe Patch algorithm: fast design of binary feedforward neural networks
8 Citations1993G Barkema, H Andree +1 more
Analytical feature extraction and spectral summation
1 Citations1996Terry Windeatt, Robert Tebbs
A formalism for analysing multilayer perceptron (MLP) networks as propagations of binary transitions along excitatory and inhibitory sensitised paths is proposed and a spectral summation is produced by characterising a Boolean function as sets of detected transitions.