Strange sparsity results about carla

[JohanvandenHeuvel]

Yeah that sounds good.

[JohanvandenHeuvel]

Hi, that should be fixed indeed. I think something like this would work (I didn't test it):

numpy.invert(numpy.isclose(delta, numpy.zeros_like(delta))) instead of delta != 0

That should fix numerical problems like that. Let me know if that works.

[aredelmeier]

Hi!

That's better but not sure it works perfectly yet. Notice that 1e-08 is still considered non-zero:

Annabelle

[JohanvandenHeuvel]

Thanks for checking! There is a tolerance parameter for numpy.isclose (https://numpy.org/doc/stable/reference/generated/numpy.isclose.html). I don't know how high the tolerance should be set.

[aredelmeier]

Setting atol=1e-5 might be a good place to start?

[JohanvandenHeuvel]

Hello,

If you want you can submit a pull request for that fix. Should be just one line (and maybe a comment), and that way everything is setup in case you would like to contribute something else as well. You might need some tools, found here.

So you just create a new branch of main, implement the fix, push the new branch to GitHub, and make a pull request for the new branch.

But I understand if you are busy with other things.

[aredelmeier]

Hi!
Sure, I can submit a pull request. Due to deadlines at the end of the week, it won't be until next week. Hope that's fine!
Annabelle

[JohanvandenHeuvel]

That's no problem at all!