Title :
Loss functions analysis in support vector regression
Author :
Popov, A.A. ; Sautin, A.S.
Abstract :
Laplace, Huber, Gauss and epsilon-insensitive loss functions in support vector regression are analyzed. Dual problem formulations and investigation results using these functions are presented. It is shown that the use of epsilon-insensitive loss function could result in biased solution.
Keywords :
regression analysis; support vector machines; Gauss loss functions analysis; Huber loss functions analysis; Laplace loss functiond analysis; epsilon-insensitive loss functions analysis; support vector regression; Equations; Least squares methods; Parameter estimation; Power system modeling; Predictive models; Probability distribution; Random variables; Robustness; Statistical analysis; System testing;
Conference_Titel :
Electronic Instrument Engineering, 2008. APEIE 2008. 9th International Conference on Actual Problems of
Conference_Location :
Novosibirsk
Print_ISBN :
978-1-4244-2825-0
DOI :
10.1109/APEIE.2008.4897179
