Repository navigation
(feature) add anisotropic meshing to LearnerND #65
Description
Activity
originally posted by Anton Akhmerov (@anton-akhmerov) at 2018-07-07T20:05:51.668Z on GitLab
Nice reference. I have thought about particle-based methods, but decided in favor of triangulation because I thought it is simpler when we don't have a luxury of being able to query the function at will.
With having the triangulation we should be able to do just that using the
transformargument to the triangulation.I am interested in implementing that. Do you have any materials to get started or some concept in mind? Should it work in a similiar way to the deviations calculation in
Learner2D?There is #141 however it's been 7 years and I don't quite remember the details 😅
Something with it didn't work well, @akhmerov or @jhoofwijk might know more?It's complicated: we ran into problems with both stability (the tesselations generated were invalid for strongly anisotropic functions) and performance. I believe to be a viable alternative to scipy, the triangulation would also likely need to be optimized to be in a compiled language.
I'm happy to chat in more detail to provide more info.
I am researching this on my own now, would be more than happy to chat about it! Would you like to move to some nicer medium than Github Issues?
(original issue on GitLab)
opened by Jorn Hoofwijk (@Jorn) at 2018-07-07T16:35:34.237Z
like this:
from this source
Still need to figure out whether de stretching should depend on the derivative or the second derivative