SVC_NU

Support Vector Classifier: NU algorithm

   This routine is outdated, use NUSVC instead

    [W,J,C] = SVC(A,TYPE,PAR,NU,MC,PD)

Input
A	Dataset
TYPE	Type of the kernel (optional; default: 'p')
PAR	Kernel parameter (optional; default: 1)
NU	Regularization parameter (0 < NU < 1): expected fraction of SV (optional; default: max(leave-one-out 1_NN error,0.05))

     MC    Do or do not data mean-centering (optional; default: 1 (to do))
     PD    Do or do not the check of the positive definiteness (optional; default: 1 (to do))

Output
W	Mapping: Support Vector Classifier
J	Object identifiers of support objects
C	Equivalent C regularization parameter of SVM-C algorithm

Description

Optimizes a support vector classifier for the dataset A by  quadratic programming. The classifier can be of one of the types  as defined by PROXM. Default is linear (TYPE = 'p', PAR = 1). In J the identifiers of the support objects in A are returned.

NU belongs to the interval (0,1). NU close to 1 allows for more class overlap.  Default NU = 0.25.

NU is bounded from above by NU_MAX = (1 - ABS(Lp-Lm)/(Lp+Lm)), where  Lp (Lm) is the number of positive (negative) samples. If NU > NU_MAX is supplied  to the routine it will be changed to the NU_MAX.

If NU is less than some NU_MIN which depends on the overlap between the classes  the algorithm will typically take a long time to converge (if at all).  So, it is advisable to set NU larger than expected overlap.

Output is rescaled in such a manner as if it were returned by SVC with the parameter C.

See also

svo_nu, svo, svc, mappings, datasets, proxm,