consider scikit-learn contribution

[clstaudt]

https://scikit-learn.org/stable/modules/generated/sklearn.svm.SVC.html

Since this implementation is famously slow, would you consider contributing your algorithm to scikit-learn?

[matzebond]

TBH this research code base probably need to much work to be included in a stable framework like sklearn. I think rebuilding the multilevel approach in python on top of the sklearn's libsvm and label_propagation would be easier and should reach most of the performance benefits.

[clstaudt]

That would count as contributing the algorithm.