Many people have tried to defend pure DRL w Nature Machine Intelligence article that actually is a hybrid model; the below youtube pointer IS to a pure convnet - but read fine print: “network works decently well for any position less than 6 moves away from solved”
#symbolphobiahttps://twitter.com/AlexRoseGames/status/1186571935611850752 …
-
So, what exactly does it mean for something to be a symbolic planning algorithm? Do we need a more nuanced terminology here? Starting with a name for methods that are more discrete and conditional than neural nets but not involving predicates and compositionality.
in The Algebraic Mind, I drew a distinction between operations over variables and structured representations, and argued separately for their presence in human cognition. eliminative connectionists tried to make do without them, relying on MLPs.
-
-
You have thought more deeply about these issues than I have, so I need to read and think more. But I think we agree that an effective (small) model of Rubik's cube dynamics requires compositional, symbolic representations (e.g., to bind colors to faces)
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.