Also, you can learn a bunch by just reading the source code. We are very picky with code organization and try to write very readable code always. Pull requests with improvements to the docstrings are welcome.
For scientific packages in general, it would be useful to provide technical references of the algorithms implemented.
Definitely. We always try to add references to the algorithms. This is one of those cases where the original paper isn’t clear. I will try to find a paper describing the method in more detail. It is considered “well-known” by a subset of the stats community.
1 Like