K

I am currently playing with Bayesian active learning where, ideally, you want to have good posteriors even when feeding labeled datums one by one, because your posterior governs your optimal...