testresults.txt.svn-base

Incrementally/decrementally update svm. very efficient!

> load nonlindata100
> svmtrain(X,y,10,5,2);
Beginning training.
Learned 50 examples.
Learned 100 examples.
Training complete!

Margin vectors:		13
Error vectors:		31
Reserve vectors:	56
Kernel evaluations:	17645

> svmloo
Beginning LOO error rate estimation.
Unlearned and tested 44 examples.
Process complete!


ans =

    18

> plot2dkm
> perturbc(50);
p = 0.00
p = 0.23
p = 0.44
p = 0.66
p = 0.84
p = 1.00
Perturbation complete!

Margin vectors:		18
Error vectors:		25
Reserve vectors:	57
Kernel evaluations:	4312

> plot2dkm
> perturbk(4);
Number of unlearned vectors: 40
Perturbation complete!

Margin vectors:		12
Error vectors:		30
Reserve vectors:	58
Kernel evaluations:	10436

> plot2dkm
> diary off