On Mar 17, 2026, at 16:40, Kostas Savvidis <kotika98@yahoo.com> wrote:
On Mar 16, 2026, at 10:21, Arnaud Becheler via Boost <boost@lists.boost.org> wrote:
why not allow users to supply a CDF estimate that actually match their actual data distribution? For Gaussian mixture, you would inject the mixture CDF.
My understanding is that this family of methods of sorting optimally relies on knowledge of inverse CDF. I am not aware of a closed form expression for inverse CDF for mixture of gaussians.
I have toyed with the idea of creating a sort routine that could do live estimation of the inverse CDF as it is traversing the list. In my mind, the difficult part is making sure that partially pre-sorted input would not produce much worse performance than unsorted.
Also, I wander what happens to performance when the data is slightly different from assumed distribution? Peter?
Cheers, Kostas
Delete that message :-) indeed you need the CDF, not inverse. I dont know what I was thinking when I was typing....