Sure, I don't disagree with that either. but I think that's quite different from the kind of deep expertise the Tal was alluding to several tweets back
-
-
"With enough data" sounds like a huge barrier, and/or disclaimer of its applicability is typical medical problems :)
-
With 1000 to 10000 datapoints, they work well. If you have less, you should probably not be mining the data, as overfit will come fast (or just use linearSVM/Ridge and stop there). https://www.sciencedirect.com/science/article/abs/pii/S1053811917305311 … Controversial, but insight from 10yr career.
- Još 1 odgovor
Novi razgovor -
-
-
Without knowing much about them, I worry that such a powerful technique, coupled with misunderstanding, might lead to even worse problems than we have now. The nice thing about conventional stats in psych is that they're incredibly simple and less powerful. But now I'm curious.
-
I think that machine learning, i.e. methods that are advertised as "intelligent" and work no matter what, have made the problem worse. It is a fallacy to think that there is *any* statistical method that can be used without actually understanding it, and expect valid results.
- Još 13 drugih odgovora
Novi razgovor -
-
-
My PhD adviser just called me out for saying basically this. I'm referencing this tweet from now on. Lol. Like of course explore other options but gradient boosted trees is a very good starting place.
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
I never thought I'd see the day...what, are you going to tell people to start sharing PNG screenshots of their jupyter notebooks for version control now?
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
Sorry we only unroll consecutive tweets from the same author, please reply to one of the main tweets from the thread; not a comment. If you think I'm buggy please DM or email
Need help? check this link: http://threadreaderapp.com/help
Kraj razgovora
Novi razgovor -
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.