CHI2 response in opencv SVM

edit

i can't reproduce it:

Ptr<ml::SVM> svm = ml::SVM::create();
svm->setKernel(ml::SVM::CHI2);

Mat_<float> data(4,2);
data << 1,1,  1.2,0.9,  3.3,2.8, 3.1,3.6;
Mat_<int> labels(4,1);
labels << -1,-1, 1,1;
svm->train(data,0,labels);

Mat_<float> query(4,2);
query << 0.8,1,  1.5,1.3,  3.9,2.5, 2.9,3.1;
Mat result;
svm->predict(query, result, 0); // use ml::StatModel::RAW_OUTPUT for the raw dot product;
cerr << result.t() << endl;

[-1,-1,1,1]

I'm trying your example.....I don't know why in your case it works perfectly....In my case, instead, I obtain something like 1.23456

the problem is in the use of ml::SVM::EPS_SVR instead of ml::SVM::C_SVC

can you specify the platform you are working on?

Linux ubuntu 16.04 and opencv3.0

Can you run OpenCV3.2 first? 3.0 is deprecated and not supported anymore ...

@alexMm1 , if the context is still your previous BOW idea - i'd think , you'd want classification(C_SVC), not regression(C_SVR) here, or not ?

(and indeed, regression returns float values, which are more like a "distance to class margin", not a label, so thresholding would not make any sense)

ok thanks @berak, understood