Explore projects

This notebook shows how autencoders can be used to identify anomalous observations using the per-observation reconstruction error. In addition, one can identify anomalous variables using the robustly standardized per-variable reconstruction error.

The notebook is a supplement to the paper "Evaluating Robustly Standardized Explainable Anomaly Detection of Implausible Variables in Cancer Data".

Various scripts I use to plot and process data from netCDF files and my implementation of two moment schemes (after ICON) with AD.