I introduced a bug into evaluate.py when I converted it to sparse matrices that only manifested when using split_components with axis=1. This is now fixed.
Additionally, contingency table generation is much faster now when a significant portion of the volume is ignored.
I introduced a bug into evaluate.py when I converted it to sparse matrices that only manifested when using split_components with
axis=1. This is now fixed.Additionally, contingency table generation is much faster now when a significant portion of the volume is ignored.