Considering that random forests have many layers and beat deep learning in most applications, maybe we just need to rebrand them as deep forests and they'll be the next big thing.
-
My wider criticism here is we should be working to expand the set of SOTA methods to cover these cases and stop looking backward. A lot of these problems might actually be better suited to deep methods using domain adaptation, self supervision, etc.
Yes, let’s stop looking backward to 50-year-old methods rebranded as deep, and invent the next ML paradigm.
-
-
At one point I actually agreed with you that traditional methods were undervalued. The last ~three years, BERT, GPT-3, seeing the massive upscaling of compute in industry changed my mind. Nobody is advocating that we ignore history, but that we not be beholden to it.
-
Most problems don’t have large datasets, unfortunately.
Keskustelun loppu
Uusi keskustelu -
Lataaminen näyttää kestävän hetken.
Twitter saattaa olla ruuhkautunut tai ongelma on muuten hetkellinen. Yritä uudelleen tai käy Twitterin tilasivulla saadaksesi lisätietoja.