Discussion of ``Concentration for (regularized) empiricalrisk minimization'' by Sara van de Geer and MartinWainwright [ MR3707417]