On Langevin updating in multilayer perceptrons

Authors:
Thorsteinn Rögnvaldsson
Affiliations:
-
Venue:
Neural Computation
Year:
1994

Citing 0
Cited 6

Dimension reduction by local principal component analysis

Neural Computation
A New Learning Algorithm Using Simultaneous Perturbation with Weight Initialization

Neural Processing Letters
Learning improvement of neural networks used in structural optimization

Advances in Engineering Software
The effects of adding noise during backpropagation training on a generalization performance

Neural Computation
WeAidU-a decision support system for myocardial perfusion images using artificial neural networks

Artificial Intelligence in Medicine
Robust training of feedforward neural networks using combined online/batch quasi-newton techniques

ICANN'12 Proceedings of the 22nd international conference on Artificial Neural Networks and Machine Learning - Volume Part II

Quantified Score

Hi-index	0.00

Visualization

Abstract

The Langevin updating rule, in which noise is added to theweights during learning, is presented and shown to improve learningon problems with initially ill-conditioned Hessians. This isparticularly important for multilayer perceptrons with many hiddenlayers, that often have ill-conditioned Hessians. In addition,Manhattan updating is shown to have a similar effect.