LearnerND triangulation incomplete

[basnijholt]

(original issue on GitLab)

opened by Bas Nijholt (@basnijholt) at 2018-10-05T07:49:51.842Z

In visualizing the 3D triangulation of the LearnerND I discovered that there must be a bug because it's incomplete.

See this notebook.

[basnijholt]

originally posted by Jorn Hoofwijk (@Jorn) at 2018-10-05T08:59:32.480Z on GitLab

And I was hoping this only happens for the anisotropic case. It may or may not be closely related to https://gitlab.kwant-project.org/qt/adaptive/merge_requests/90#note_19443

Also I do really like plotly, last week I've also been playing around with it

[basnijholt]

originally posted by Jorn Hoofwijk (@Jorn) at 2018-10-05T10:03:22.450Z on GitLab

I can actually confirm this is exactly what happens in https://gitlab.kwant-project.org/qt/adaptive/merge_requests/90#note_19443

the simplex connected to the vertex in the top-right overlaps with other simplices, while simplices should actually never overlap

I have actually never observed this apart from the anisotropic learner, so I am partly surprised and worried and relieved at the same time.

I am certain I can fix this, but I am unsure how long it will take, but I will make sure it gets fixed.

[basnijholt]

originally posted by Jorn Hoofwijk (@Jorn) at 2018-10-07T08:14:08.781Z on GitLab

I noticed there is some error in the visualisation as well, using this notebook, and replacing the runner with a simple runner:

adaptive.runner.simple(learner, goal=lambda l: l.npoints >= 5)

I saw that not all edges are plotted while they are present. I will rewrite the notebook a little to fix it, and to see if the missing/overlapping simplices still occur, (i think they might)

[basnijholt]

originally posted by Jorn Hoofwijk (@Jorn) at 2018-10-08T06:10:02.184Z on GitLab

Using the following notebook, this is no longer visible: LearnerND_plotly_110.ipynb

I think showing the triangulation is indeed very effective for debugging and for example finding out what goes wrong in what cases in gitlab:!90.

I will take back what I said here: https://gitlab.kwant-project.org/qt/adaptive/issues/110#note_19620 It was in fact not the same thing as in https://gitlab.kwant-project.org/qt/adaptive/merge_requests/90#note_19443

I also think this issue may be closed.