Fitting 3D Data Points By Extending The Neural Networks Paradigm
Price
Free (open access)
Transaction
Volume
30
Pages
10
Published
2001
Size
1,070 kb
Paper DOI
10.2495/CMEM010791
Copyright
WIT Press
Author(s)
A. Iglesias, A. Galvez
Abstract
Fitting 3D data points by extending the neural networks paradigm A. Iglesias, A. Galvez Dept. of Applied Mathematics and Computational Sciences, University of Cantabria, Spain Abstract This paper describes a new method to fit 3D data points from engineer- ing environments (for example, from a numerical-controlled machine or by digitizing a real model). Since these points are subjected to measurement errors, approximation techniques are required. Among them, several tech- niques based on neural networks have been extensively applied in the last recent years. However, for data points from a B-spline surface the approx- imation problem could not be well described in terms of neural networks. To overcome this limitation, this paper introduces a new extension of the neural networks paradigm, the functional networks, in which the weights are replaced by neural functions with a multivariate character. In addition, these neural functions can be different for different neurons. The perfor-
Keywords