search

TeaspoonTDA / teaspoon

Topological Signal Processing in Python
https://teaspoontda.github.io/teaspoon/
GNU General Public License v3.0
23 stars 9 forks source link

Wasserstein distance using optimal transport not showing up in the documentation #6

Closed lizliz closed 1 year ago

lizliz commented 2 years ago

We have a command for computing wasserstein distance using the optimal transport (ot) package. However, it doesn't show up in the documentation automatically.

I haven't done a lot of testing and I can't remember if it wrote it as such, but I believe this will do any $p$ and $\infty$ (so bottleneck) distance. If this is the case, we could remove dependency on Hera entirely.

lizliz commented 2 years ago

Forgot to mention this is the command wassersteinDist in teaspoon.TDA.Distance

barnesd8 commented 1 year ago

I think we can remove Hera and either update using the ot package, or include the bottleneck from persim, but it can be an option in the wassersteinDist function. Currently when setting q = np.inf, it computes chebyshev distance, but I am unable to get it to match the results from the persim bottleneck library or the prior example from Hera as well.

Screenshot 2023-01-02 at 2 07 25 PM

I suspect the matching for computation of bottleneck is being done differently. Screenshot attached.

barnesd8 commented 1 year ago

This is resolved in the documentation and will use the persim package for bottleneck distance