The conception of the paper was more than 3 years old, but I have been quite busy with my company and took me so long to write it and revise it a few times. I have several more to write in the pipeline...
-
-
Prikaži ovu nitHvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
very nice overview. daw and i proposed a role of mental modeling in behavior controlled by emtional experience. looking forward to reading your parer more thoroughly
-
Thank you! In the current version, we didn't have much idea about emotional experience. It would be great if we could have a discussion.
- Još 2 druga odgovora
Novi razgovor -
-
-
Cool paper. I didn't get exactly why "the bottom-up mes- sage passing of prediction errors corresponds to encoding (or data compression)". Why is the generation of prediction errors a form of data compression? Not exactly sure how VAEs correspond to predictive coding networks.
-
Thanks for the kind comment. Let me think of a way to explain how predictive coding is similar to (but not exactly the same as) autoencoders. Perhaps, if you consider just three layers (input-latent-output), maybe the similarity is more apparent?
- Još 3 druga odgovora
Novi razgovor -
-
-
Nicely done. But how does this relate to a previous paper on Information Closure?
-
That's a topic for the next paper. We are trying to work out what implies what (e.g. does FEP imply IC?).
- Još 2 druga odgovora
Novi razgovor -
-
-
Nice piece, congratulations! Made me realise how we have now at least three different types of "generative models" in the literature though (w/ statistical & mechanistic ones)...
#WordsCollision#LumperSplitter cc@dngmanHvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
I like this view. But it does imply that what is mostly studied in our field - consciousness of the CURRENT sensory input - is an epiphenomenon, which emerged as a consequence of generating alternative representations. Should the focus of the field shift toward memory/planning?
-
Current sensory input is always preceded by memory/expectations so they aren’t really separable. It’s always a problem of inference.
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.