A=ONE_VS_REST(C,H) returns an one_vs_rest object which trains algorithm
C on sub problems of type class i versus all other classes,
and is initialized with hyperparameters H. The classifiers are combined
by outputting the class with the largest positive output (this assumes
that classifiers output real values indicating confidence rather than
just +1,-1)
Model
child=svm -- classifier to use for each sub-problem
Methods:
train, test, get_w
Example:
c1=[-1,1];c2=[1,1];c3=[0,-1];
X1= randn(50,2)+repmat(c1,50,1);
X2= randn(50,2)+repmat(c2,50,1);
X3= randn(50,2)+repmat(c3,50,1);
note the class label format!
Y1= [ones(50,1),-ones(50,1),-ones(50,1)];
Y2= [-ones(50,1),ones(50,1),-ones(50,1)];
Y3= [-ones(50,1),-ones(50,1),ones(50,1)];
d=data([X1;X2;X3],[Y1;Y2;Y3]);
[r,a]=train(one_vs_rest(svm(kernel('rbf',2))),d)
Test class centers
dtest=data([c1;c2;c3]);
rtest=test(a,dtest)
| |
Reference : Multi-class Support Vector Machines |
| |
Author : Jason Weston , C. Watkins |