A cosine-modulated Gaussian activation function for hyper-hill neural networks

Lee, S.-W.; Moraga, Claudio

doi:10.1109/icsigp.1996.566581

Cited by 3 publications

(7 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The authors compare the hybrid structure in a multifrequency signal classification problem, concluding that even if the combination of the three activation functions performs better than the sigmoid (in terms of convergence speed) and the Gaussian (in terms of noise rejection), the sinusoidal activation function by itself still achieves better results. Another work investigating an activation function based on sinusoidal modulation can be found in [25], where the authors propose a cosine modulated Gaussian function. The use of sinusoidal activation function is deeply investigated in [26], where the authors present a comprehensive comparison between eight different activation functions on eight different problems.…”

Section: Activation Functions For Easy Trainingmentioning

confidence: 99%

On Training Efficiency and Computational Costs of a Feed Forward Neural Network: A Review

Laudani

Lozito

Fulginei

et al. 2015

Computational Intelligence and Neuroscience

View full text Add to dashboard Cite

A comprehensive review on the problem of choosing a suitable activation function for the hidden layer of a feed forward neural network has been widely investigated. Since the nonlinear component of a neural network is the main contributor to the network mapping capabilities, the different choices that may lead to enhanced performances, in terms of training, generalization, or computational costs, are analyzed, both in general-purpose and in embedded computing environments. Finally, a strategy to convert a network configuration between different activation functions without altering the network mapping capabilities will be presented.

show abstract

Section: Activation Functions For Easy Trainingmentioning

confidence: 99%

On Training Efficiency and Computational Costs of a Feed Forward Neural Network: A Review

Laudani

Lozito

Fulginei

et al. 2015

Computational Intelligence and Neuroscience

View full text Add to dashboard Cite

show abstract

“…Sometimes different activation functions [4][5][6][7][8][9][10][11][12][13][14] are acquired for different networks so that it resulting in better performances. An activation function or transfer functions for the hidden nodes in MLP are needed to introduce nonlinearity into the network.…”

Section: Activation Functionsmentioning

confidence: 99%

“…In [14], a Cosine-Modulated Gaussian activation function for Hyper-Hill neural networks has been proposed. The study compared the Cosine-Modulated Gaussian, hyperbolic tangent, sigmoid and symsigmoid function in cascade correlation network to solve sonar benchmark problem.…”

Section: Introductionmentioning

confidence: 99%

“…Some researchers had also proposed special activation function such as logarithmic [4], Adaptive Spline [5], Type-2 Fuzzy [7], Elementary Transcendental [8], Hermite Polynomial [9], periodic and monotonic [11], Novel Adaptive [12] and Cosine-modulated Gaussian activation function [14]. However, the most recommended activation function among neural network researchers for MLP network is sigmoid or hyperbolic tangent [6], [10].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Suitable MLP Network Activation Functions for Breast Cancer and Thyroid Disease Detection

Isa

Saad

Omar

et al. 2010

2010 Second International Conference on Computational Intelligence, Modelling and Simulation

View full text Add to dashboard Cite

“…The experimental results show that the proposed model has better performance than conventional MLPs. A study conducted by S. W. Lee and C. Moraga [6] propose the function of Cosine-Modulated Gaussian for Hyper-Hill neural networks. The study has compared the Cosine-Modulated Gaussian, hyperbolic tangent, sigmoid and symsigmoid function in cascade correlation network to solve sonar benchmark problem.…”

Section: Introductionmentioning

confidence: 99%

Comparisons of MLP transfer functions for different classification classes

Isa

Fauzi

Sharif

et al. 2012

2012 IEEE International Conference on Control System, Computing and Engineering

View full text Add to dashboard Cite

This paper presents a comparison study of two different MLP transfer functions for three different classification cases of breast cancer, thyroid disease and weather classification. The transfer functions under investigation are sigmoid and hyperbolic tangent. In the study, MLP network was trained and tested to investigate the ability of the network to classify the breast cancer correctly between benign cell and malignant cell, classifying thyroid disease into normal, hyper or hypo thyroid and classifying weather conditions into four types; rain, cloudy, dry day and storm. Levenberg-Marquardt algorithm is adopted to train MLP network since it is the fastest training and ensure the best converges towards a minimum error. The performance of MLP networks was evaluated in terms of percentages for correct classification of the target outputs. Both functions are able to give accuracies up to 99% for classifying correctly. The hyperbolic tangent function had shown the capability of achieving the highest accuracy of an MLP performance with less number of hidden nodes.

show abstract

A cosine-modulated Gaussian activation function for hyper-hill neural networks

Cited by 3 publications

References 10 publications

On Training Efficiency and Computational Costs of a Feed Forward Neural Network: A Review

On Training Efficiency and Computational Costs of a Feed Forward Neural Network: A Review

Suitable MLP Network Activation Functions for Breast Cancer and Thyroid Disease Detection

Comparisons of MLP transfer functions for different classification classes

Contact Info

Product

Resources

About