non gradient based DLhttps://twitter.com/GaryMarcus/status/1080549287212314626 …
-
http://www.jmlr.org/proceedings/papers/v48/taylor16.pdf … not sure if this solves problems but there are already approaches w/o gradients. Just as GANs dont always necessarily mean cnn's, training may be without gradient based optimization.
It’s interesting. @ylecun, as a person committed to gradients, curious your take on the link above. (And also whether you would still count it as deep learning.)
-
-
I don't think Prof Lecun is committed to gradientshttps://github.com/facebookresearch/nevergrad …
-
And for non-gradient based opt on DNNs, stay tuned, my personal work is in progress
End of conversation
New conversation -
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.