That's an interesting discussion! Having read the Vovk papers, this blog post definitely presents things much more clearly. The original papers often don't adhe

That's an interesting discussion! Having read the Vovk papers, this blog post definitely presents things much more clearly. The original papers often don't adhere to the standard definition/lemma/proof style of mathematical exposition, which makes them really hard to follow. It's also an interesting coincidence that this story on the front page today. I'm giving a talk tomorrow at AAAI on some work that extends this theory. We show how to do uncertainty estimation (e.g. calibrated probabilities for ML classifiers) under fully adversarial assumptions (input data can be chosen by an adversary). I'll do a shameless plug and post the paper here, in case people are interested in this general topic: https://arxiv.org/abs/1607.03594 https://arxiv.org/abs/1607.03594