


PSVC Proximal Support Vector Classification
===============================================================================
Usage: [w,gamma,trainCorr, testCorr,cpu_time,nu]=psvc(A,d,k,nu,output,bal)
A and d are both required, everything else has a default
An example: [w gamma train test time nu] = psvm(A,d,10);
Input:
A is a matrix containing m data in n dimensions each.
d is a m dimensional vector of 1's or -1's containing
the corresponding labels for each example in A.
k is k-fold for correctness purpose
nu - the weighting factor.
-1 - easy estimation
0 - hard estimation
any other value - used as nu by the algorithm
default - 0
output - indicates whether you want output
If the input parameter bal is 1
the algorithm weighs the classes depending on the
number of points in each class and balance them.
It is useful when the number of point in each class
is very unbalanced.
Output:
w,gamma are the values defining the separating
Hyperplane w'x-gamma=0 such that:
w'x-gamma>0 => x belongs to A+
w'x-gamma<0 => x belongs to A-
w'x-gamma=0 => x can belongs to both classes
nu - the estimated or specified value of nu
For details refer to the paper:
"Proximal Support Vector Machine Classifiers"
available at: www.cs.wisc.edu/~gfung
For questions or suggestions, please email:
Glenn Fung, gfung@cs.wisc.edu
Sept 2001.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%