Seems easy? Now do this <l2st> in thousands of dimensions with complicated distributions and overlap between anomalies and normal points </l2st>