DocumentCode :
3548718
Title :
A study on the differences in the interpolation capabilities of models
Author :
Juutilainen, Ilmari ; Röning, Juha ; Laurinen, Perttu
Author_Institution :
Intelligent Syst. Group, Oulu Univ., Finland
fYear :
2005
fDate :
28-30 June 2005
Firstpage :
202
Lastpage :
207
Abstract :
We examined the interpolation capabilities of learning methods using simulated data sets and a real data set. We compared five common learning methods for their generalisation capability on the boundaries of the training data set also; we examined the effects of the complexity of models on interpolation capability. Our main results were that there are differences between the different model families, but model complexity does not have a major effect on interpolation capability. The multi-layer perceptron, support vector regression and additive spline models outperformed local linear regression and quadratic regression in interpolation capabilities. Information about the interpolation capability of models is useful when, for example, evaluating the reliability of prediction.
Keywords :
generalisation (artificial intelligence); interpolation; learning (artificial intelligence); multilayer perceptrons; regression analysis; splines (mathematics); support vector machines; additive spline model; generalisation; interpolation; learning method; model complexity; multi-layer perceptron; outperformed local linear regression; quadratic regression; real data set; simulated data set; support vector regression; Data engineering; Industrial training; Intelligent systems; Interpolation; Learning systems; Predictive models; Spline; Statistical learning; Testing; Training data;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Soft Computing in Industrial Applications, 2005. SMCia/05. Proceedings of the 2005 IEEE Mid-Summer Workshop on
Print_ISBN :
0-7803-8942-5
Type :
conf
DOI :
10.1109/SMCIA.2005.1466973
Filename :
1466973
Link To Document :
بازگشت