Prediction for you, @marius10p, is this: the distribution of activity, a, during motor action x should be roughly p(a|v(x)) where the function v(x) is the expected value of visual inputs in near future given current motor commands.
-
-
Prikaži ovu nit
-
Key would be for you to have an experimental prep (and trained animals) where you can say with reasonable confidence what that v(x) fn is.
Prikaži ovu nit
Kraj razgovora
Novi razgovor -
-
-
Yeah sensory parts of the brain aren't just waiting for stimuli...it's more like throwing another ball to someone who is always juggling.
- Još 4 druga odgovora
Novi razgovor -
-
-
The neurons are representing actions, not the sensory consequences of those actions. The latter would have looked very different. For example in darkness, the sensory consequences of movement are still darkness, so there should be no response! Read the paper for the fine details.
-
Maybe.. you 100% know what the correct response should be, but do the neurons? I will read the paper before further pronouncements though.
Kraj razgovora
Novi razgovor -
-
-
Here's is some of the most direct evidence that this is the casehttp://science.sciencemag.org/content/331/6013/83.long …
-
I really like this paper, but the lack of relation to behaviour and time always struck me as potentially problematic. It assumes we try to build a generative model of sensory snapshots, but more likely, we try to build model of sensory stimuli conditioned on actions & past.
Kraj razgovora
Novi razgovor -
-
-
Now you're "sampling" Schmidhuber:https://twitter.com/hardmaru/status/978794349101137920 …
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.
Can neural network agents dream, and learn inside of their own dream worlds?
Read more to find out:
Full Interactive Article: