This is why multi-resolution-analysis ala wavelets is extremely powerful, even with set basis functions.
It's an under-appreciated fact, but learning hierarchical-modular representations (DL) is not tied to NNs, differentiability, or SGD
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
By tied to, do you mean exclusive to?
-
Yes, deep learning != NNs
End of conversation
New conversation -
-
-
Ehhh, but differentiable NNs are so good at it. Ever mess around with LatentSVM? I feel like the difference is night and day
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
So much to learn. Is representation learning limited to NN's? Or what is passed through other architecture are learned representations too?
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.