Anomalies are data examples that are, simply put, weird. They are far from typical data points and make us wonder if they are not generated by a completely different process. I am interested in them as a mechanism for scientific discovery: weird, anomalous examples test and often break our current theories.

Robust random cut forests and isolation forests are two approaches to detecting anomalies in data. Since they are related models, I provide a combined code base that allows one to use them in conjunction. See the GitHub repo for more details. A description of the ideas is found in the associated write-up.