eloquentarduino/micromlgen

Decision function computed doesn't match with sklearn.svm.SVC

aravindsiv opened this issue · 1 comments

Hi,

I train my SVC in Python and port it to C++ as follows:

with open(fname, "w") as f: f.write(port(clf))

However, the decision function computed by clf.decision_function(X) (in Python) and the one computed inside Eloquent::ML::Port::SVM::predict (in the generated header file) do not match. The predictions do not match.

Is there something I'm missing?