link is broken
-
-
-
sjould work now again
Kraj razgovora
Novi razgovor -
-
-
There's some problem with the link. Can you please provide another way to access your note?
-
should work now again
Kraj razgovora
Novi razgovor -
-
-
I’m eager to read this but the link is broken.
-
shout work now again
- Još 1 odgovor
Novi razgovor -
-
-
But recent papers on disentanglement using generative models show that disentanglement doesn't improve sample efficiency. And generative models don't know what they don't know.
-
It is discriminative models that don't know what they don't know. Generative models capture P(x|y) and P(y) from which we can compute P(x). When we see an x for which P(x) is small, we know that we don't know how to handle it.
- Još 11 drugih odgovora
Novi razgovor -
-
-
Nice! Generative modeling is the way forward, but we are stuck in many situations where discrim models still work better. A nice trick we found is to invert the *discim* model in service of a larger generative model: http://slanglab.cs.umass.edu/doc_prevalence/
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
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.