benchmarking the linear kernel in SVC

[amueller]

Using kernel='linear' in SVC basically makes no sense. You should be using LinearSVC and probably set dual=False.

[bibikar]

DAAL will only work in the dual space for SVM, so daal4py's patches patch SVC instead of LinearSVC. We could possibly add an option to benchmark LinearSVC instead, or only choose SVC by default if DAAL is present. What do you think?

[amueller]

The point is not what DAAL does but whether the benchmark is fair. If you want to compare solving a linear SVM between daal and scikit-learn, you should be using LinearSVC on the sklearn side.