Neither pre-winter symbolic nor connectionist stuff has all that much philosophically in common with how people are building new architectures. Probably better to use a new term anyhow, make room for new syntheses of the wealth of empirical evidence available now.
Didn’t say you couldn’t enhance deep learning w external memory etc, just called for truth-in-labeling, and clarity about what components matter instead of a coarse, imperialist use of term. Why not call deep learning + load/store/memory “symbols + DL”?https://twitter.com/zergylord/status/955119426571038720 …
-
-
-
People were looking to extend neural networks to richer models of computation in the early 1990s too. https://arxiv.org/pdf/1711.05738.pdf …
- 11 more replies
New conversation -
-
-
Just shared a talk from Indiana - adds some new stuff my students doing to approach DL + symbols (and explain why) …https://www.slideshare.net/jahendler/kr-in-the-age-of-deep-learnign …
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Totally agree that your convention would add clarity, but deep learning *as practiced* has included these things for a few years. External memory is an architectural change; same optimization and objectives. Such changes have been viewed as core to the field.
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.