POPS PAC Ellipse

Coming soon!

Formally the POPS method does not have a robust learning scheme at small N.

This can be rectified by casting the generalization error in a form suitable for PAC-Bayes analysis.