The No Free Lunch theorem is often misinterpreted. There are several universal approximation bounds (Cybenko, Barron, Vapnik) for ML model classes. These show that any reasonably behaved function can be learned, but the price of lunch is huge amounts of data
Exactly, the ambition of ML often seems to be free lunch. And the No Free Lunch theorem says that will never happen.https://twitter.com/gewaltig/status/1193185916539621376 …
-
-
-
In other words, I agree with
@GaryMarcus. I just wanted to dampen peoples' enthusiasm for citing the No Free Lunch theorem in these arguments. - 2 more replies
New conversation -
-
-
Sitting next to David Wolpert right now at a conference.
-
never met, but tell him I am a fan :)
End of conversation
New conversation -
-
-
Please. Of course, we want a free lunch. We want lots of free lunches. Otherwise we'll still be at it when the sun goes supernova. Babies are not born knowing how to speak and walk. They use the free lunches of unsupervised perceptual and motor learning to do those things.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.