Interesting point of view. Will neural networks (or other state-of-the-art methods) end up refuting Occam’s razor, or will they turn out to be “simple” after all?https://twitter.com/pmddomingos/status/1392321198566961155 …
You can make anything you want simple a posteriori by changing the representation (indeed, the history of science is full of this), but then "simplicity" is vacuous.
-
-
I was thinking the same thing as I Tweeted it, but maybe there’s some global sense in which all these algorithms are “simple” that we haven’t discovered yet. Or is that too much to hope for?
-
There's no universal notion of simplicity. Better to think in terms of hypothesis space size (or prior entropy), but even that is only an upper bound on the amount of search you do. See: https://homes.cs.washington.edu/~pedrod/papers/ijcai99.pdf …
- Näytä vastaukset
Uusi keskustelu -
-
-
Yes, but sometimes changing representation to make one thing simpler makes other things simpler too. Those are the big wins.
Kiitos. Käytämme tätä aikajanasi parantamiseen. KumoaKumoa
-
Lataaminen näyttää kestävän hetken.
Twitter saattaa olla ruuhkautunut tai ongelma on muuten hetkellinen. Yritä uudelleen tai käy Twitterin tilasivulla saadaksesi lisätietoja.