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