I am doing a project about svm and need to find out the support vectors of the data set. But I do not know what the exact criteria are for support vectors in scikit-learn. Do anyone know how to find the exact mathmatical funtions or source code behind it?
Here a little bit of the math. The references at the end have more detail.
The implementation is here.
In scikit-learn, the Support Vector Machine (SVM) implementation uses the LIBSVM library, which is written in C++. While the exact mathematical functions and source code for the underlying LIBSVM library are not directly accessible within scikit-learn, you can still access the support vectors and their properties using the scikit-learn API.
To retrieve the support vectors from an SVM model in scikit-learn, you can use the support_vectors_ attribute of the trained SVM model object. Here’s an example:
from sklearn import svm
Create and train the SVM model
X = [[0, 0], [1, 1]]
y = [0, 1]
clf = svm.SVC()
clf.fit(X, y)
Access the support vectors
support_vectors = clf.support_vectors_
print(support_vectors)
The support_vectors_ attribute will provide you with an array of the support vectors found during the training process. Each support vector is a point from the original dataset that lies on or within the margins defined by the SVM model.