After displaying a simple (bivariate) example of the calculation of Mahalanobis distances, the notebook illustrates their efficiency for identifying outliers, first using a classical covariance determination (from numpy), then using a robust covariance estimate (from scikit-learn). Finally, the combined use of the scikit-learn EllipticEnvelope and DecisionBoundaryDisplay, which achieves similar results on a (very) simple dataset, is shown for illustration purposes.
A notebook querying Gaia astronomical data for celestial objects located in the vicinity of a nearby double star cluster, aiming at identifying members of each of the clusters constituting the pair.