SVM: do we need the polynomial kernel

[mfeurer]

in initial experiments it is way slower than the RBF kernel.

[VladimirGl]

It also sometimes causes hangs for me then use ask on multithread mode (so I have separate model fitters and one ensembler).
So it was killed by time limits inside models fitters, but then I call refit on ensembler thread on whole data on, or on different folds (if use external cv) it never stops (may be related to some data scaling issues, not sure). I just set max number of iterations to couple of thousands instead of -1 to avoid this.

[mfeurer]

Another related question: do we need to apply calibration via Platt Scaling to SVMs? Getting the most out of ensemble selection suggests that this is not necessary, so the time required to calibrate the SVM might rather be spent on searching for better hyperparameters.

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs for the next 7 days. Thank you for your contributions.

[mfeurer]

Closing, see #1314 for the reasoning.