Another related problem that might be worth talking about is overfitting. Deciding what is noise and what isn’t noise and the related problem of how finely grained categories should be can be tough.
Yes! Whole new issue!
Multi-scale methods can help (I like this example a lot) but I think there’s still a little ad-hoc stuff hidden within the algorithm.
Another related problem that might be worth talking about is overfitting. Deciding what is noise and what isn’t noise and the related problem of how finely grained categories should be can be tough.
Yes! Whole new issue! Multi-scale methods can help (I like this example a lot) but I think there’s still a little ad-hoc stuff hidden within the algorithm.