v1.2.2

Add clarification on the implication of bounds choices in the main pruning functions.

v1.2.1

Add experiment showing modularity is not K-monotonic (the number of communities may actually temporarily decrease as the resolution parameter, gamma, increases).

Also increase the granularity of the scikit-learn dependency version in order to maintain support for Python 3.7. Many packages have dropped anything below 3.8, but I'm trying to not drop it so far ahead of its end-of-life date (near the end of June 2023) if possible.

v1.2.0

Remove upstream champ usage and re-enable Python 3.7/3.10 testing.

Roughly speaking, this makes all the champ_usage tests ~10% faster on average and eliminates multiple warnings from the upstream champ code (numerical optimization issues, deprecation warnings, etc.).

It also makes our CHAMP code more robust and eliminates some non-deterministic failures that were observed in the upstream code. These used to cause failures in at least one CHAMP test case occasionally.

We keep the champ dependency where possible (i.e. for Python <3.10) just in case the user wants to run some of the related experiments.

Full Changelog: 1.1.3...1.2.0

v1.1.3

Add preprint citation and tweak testing and dependency requirements.

v1.1.2

Significantly expand documentation and make minor tweaks to experiments and functionality.

v1.1.0

Add documentation and increment version number