AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
How Do We Get the Best Out of Difficult Examples?
Some of the first things i'd come up with would be ideas like focal loss, for example, in retina. How do we figure out which examples are pretty difficult for us? And then once we've determined what type of difficult example we're dealing with, how do we then act on that information going forward? Thias: There's been a lot of really interesting work which has focused on isolations o like learning a typical examples upweighting them.