robustlssvm

Purpose

Robust training in the case of non-Gaussian noise or outliers

Basic syntax

>> [alpha, b] = robustlssvm({X,Y,type,gam,sig2,kernel})
>>  model     = robustlssvm(model)

Robustness towards outliers can be achieved by reducing the influence of support values corresponding to large errors.

Full syntax

• Using the functional interface:

  >> [alpha, b] = robustlssvm({X,Y,type,gam,sig2})
  >> [alpha, b] = robustlssvm({X,Y,type,gam,sig2,kernel})
  >> [alpha, b] = robustlssvm({X,Y,type,gam,sig2,kernel, preprocess})
  >> [alpha, b] = robustlssvm({X,Y,type,gam,sig2,kernel, preprocess}, {alpha,b})
  

  Outputs
  	alpha	N$\times$1 matrix with support values of the robust LS-SVM
  	b	1$\times$1 vector with bias term(s) of the robust LS-SVM
  Inputs
  	X	N$\times$d matrix with the inputs of the training data
  	Y	N$\times$1 vector with the outputs of the training data
  	type	'function estimation' ('f') or 'classifier' ('c')
  	gam	Regularization parameter
  	sig2	Kernel parameter(s) (for linear kernel, use [])
  	kernel(*)	Kernel type (by default 'RBF_kernel')
  	preprocess(*)	'preprocess'(*) or 'original'
  	alpha(*)	Support values obtained from training
  	b(*)	Bias term obtained from training

• Using the object oriented interface:

  >> model = robustlssvm(model)
  

  Outputs
  	model	Robustly trained object oriented representation of the LS-SVM model
  Inputs
  	model	Object oriented representation of the LS-SVM model

See also:

trainlssvm, tunelssvm, rcrossvalidate