LS-SVM hyperparameter selection with a nonparametric noise estimator

Lendasse, Amaury;Ji, Jongnan;Reyhani, Nima;Verleysen, Michel
(2005) 15th International Conference on Artificial Neural Networks (ICANN 2005) — Location: Warsaw (Poland) (11.September.2005)

Files

C27-LS-SVMHyperparameterSelectionwithaNonparametricNoiseEstimator.pdf
  • Restricted Access
  • Adobe PDF
  • 384.35 KB
Request a copy

Details

Authors
  • Lendasse, AmauryHelsinki University of Technology
    Author
  • Ji, JongnanHelsinki University of Technology
    Author
  • Reyhani, NimaHelsinki University of Technology
    Author
  • Verleysen, MichelUCLouvain
    Author
Abstract
This paper presents a new method for the selection of the two hyperparameters of Least Squares Support Vector Machine (LS-SVM) approximators with Gaussian Kernels. The two hyperparameters are the width sigma of the Gaussian kernels and the regularization parameter lambda. For different values of sigma, a Nonparametric Noise Estimator (NNE) is introduced to estimate the variance of the noise on the outputs. The NNE allows the determination of the best lambda for each given sigma. A Leave-one-out methodology is then applied to select the best sigma. Therefore, this method transforms the double optimization problem into a single optimization one. The method is tested on 2 problems: a toy example and the Pumadyn regression Benchmark.
Affiliations

Citations

  • APA
  • Chicago
  • FWB

Lendasse, A., Ji, J., Reyhani, N., & Verleysen, M. (2005). LS-SVM hyperparameter selection with a nonparametric noise estimator. Lecture Notes in Computer Science, 3697, 625-630. https://hdl.handle.net/2078.5/254011 (Original work published 2005)