Problem
I have encountered a situation where the KDE calculation in violinplot can be significantly improved by using precomputed weights (memory overflow crash vs several seconds).
It would help me significantly (I had to write my own implementation) if it was exposed.
Proposed solution
Expose the weights parameter of the scipy.stats.gaussian_kde in violinplot function.
Problem
I have encountered a situation where the KDE calculation in
violinplotcan be significantly improved by using precomputed weights (memory overflow crash vs several seconds).It would help me significantly (I had to write my own implementation) if it was exposed.
Proposed solution
Expose the weights parameter of the
scipy.stats.gaussian_kdeinviolinplotfunction.