OAOSVM |
Multi-class SVM using One-Against-One decomposition.
Synopsis:
model = oaosvm( data )
model = oaosvm( data, options )
Description:
model = oaosvm( data ) uses one-agains-one deconposition
to train the multi-class Support Vector Machines (SVM)
classifier. The classification into nclass classes
is decomposed into nrule = (nclass-1)*nclass/2 binary
problems.
model = oaosvm( data, options) allows to specify the
binary SVM solver and its paramaters.
Input:
data [struct] Training data:
.X [dim x num_data] Training vectors.
.y [1 x num_data] Labels of training data (1,2,...,nclass).
options [struct] Control parameters:
.bin_svm [string] Function which implements the binary SVM
solver; (default 'smo').
.verb [1x1] If 1 then a progress info is displayed (default 0).
The other fields of options specifies the options of the binary
solver (e.g., ker, arg, C). See help of the selected solver.
Output:
model [struct] Multi-class SVM majority voting classifier:
.Alpha [nsv x nrule] Weights (Lagrangeans).
.bin_y [2 x nrule] Translation between binary responses of
the discriminant functions and class labels.
.b [nrule x 1] Biases of discriminant functions.
.sv.X [dim x nsv] Support vectors.
.nsv [1x1] Number of support vectors.
.trnerr [1x1] Training error.
.kercnt [1x1] Number of kernel evaluations.
.options [struct[ Copy of input argument options.
Example:
data = load('pentagon');
options = struct('ker','rbf','arg',1,'C',1000,'verb',1);
model = oaosvm( data, options );
figure;
ppatterns(data); ppatterns(model.sv.X,'ok',13);
pboundary( model );
See also
MVSVMCLASS, OAASVM.
About: Statistical Pattern Recognition Toolbox
(C) 1999-2005, Written by Vojtech Franc and Vaclav Hlavac
Czech Technical University Prague
Faculty of Electrical Engineering
Center for Machine Perception
Modifications:
25-jan-2005, VF, option solver replaced by bin_svm
26-may-2004, VF
4-feb-2004, VF
9-Feb-2003, VF