FLDQP |
Fisher Linear Discriminat using Quadratic Programming.
Synopsis:
model = fldqp( data )
Description:
This function computes the binary linear classifier based
on the Fisher Linear Discriminant (FLD) using the Quadratic
Programming (quadprog) optimization. The inputs are
binary labeled training vectors. The parameter vector W
of the linear classifier
q(x) = 1 for W'*x + b >= 0
= 2 for W'*x + b < 0
is computed to maximize class separability criterion.
The bias b is determined to lie between means of training
data projected onto direction W.
Input:
data [struct] Binary labeled training vectors.
.X [dim x num_data] Training vectors.
.y [1 x num_data] Labels (1 or 2).
Output:
model [struct] Binary linear classifier:
.W [dim x 1] Parameter vector the linear classifier.
.b [1x1] Bias of the linear classifier.
.separab [1x1] Meassure of class separability.
Example:
trn = load('riply_trn');
tst = load('riply_tst');
model = fldqp( trn );
ypred = linclass( tst.X, model);
cerror(ypred, tst.y)
figure; ppatterns(trn); pline(model);
See also
FLD, LINCLASS.
About: Statistical Pattern Recognition Toolbox
(C) 1999-2003, Written by Vojtech Franc and Vaclav Hlavac
Czech Technical University Prague
Faculty of Electrical Engineering
Center for Machine Perception
Modifications:
21-may-2004, VF
1-may-2004, VF
30-apr-2004, VF
24-Feb-2003, VF
1-Feb-2003, VF