To me this post just underscored the need for priors (as the author says in the "Looking to the Future" section) For a neuro person this is stating the obvious but priors are encoded in multiple ways in the brain: genetically, developmentally, culturally
“Deep Reinforcement Learning Doesn't Work Yet” - outstanding blog from February by @alexirpan that I somehow missed. DRL experts: have these problems been addressed since?http://www.alexirpan.com/2018/02/14/rl-hard.html …
-
-
-
So DRL will never "work" if "working" means learning from zero experience with no priors. I mean, I guess you already agree with this. Why doesn't the DRL community?
- 2 more replies
New conversation -
-
-
in 9 months? Snigger. There is Bonsai though...
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
interesting post, thanks for sharing. loved this quotehttps://twitter.com/samim/status/1065620014118653952 …
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
These are all intractable DRL characteristics; there shouldn't be an intimation we can fix any outside of hybridization...I love his list "so far" of what works in DRL; it's an inch long, if you put each of the seven things he could think of on its own line.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Easily the best description of the limitations DL that I’ve seen. Not unnecessarily arcane, and covers all the main points.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.