I finally get to release anomix. This is a mixture modeling package for anomaly detection that I had been working on for several months over the past year. Originally developed for a larger project but never implemented, I nurtured it in spare time, and was able to separate it into a stand alone repository for anomaly detection on univariate data.
It supports several distributions and implements MLE estimation where that is well defined, and various other, non-MLE estimation approaches where it is not (looking at you, Cauchy, Zipf, and t-distribution).
It is not massively useful, I will admit, but it is pretty handy for its few use cases. It is also a code project meant to reflect maturation as a developer and data scientist — I have open sourced a general use statistical machine learning package on PyPI.
Install it with pip install anomix, or clone and build the repo from source.
Again, thank you to my employer TRSS for allowing me to develop and release this project. I hope people use it. See also the xView3 challenge and the hashtag community detection paper, from the same period.