Re:Why deep learning works even though it shouldn’t
by dougabug on Tuesday October 20, 2020 @02:16PM
(#24840635)
ReLU networks are piecewise linear.
[ Parent
| on HN ]
Re:Why deep learning works even though it shouldn’t
by RandyRanderson on Tuesday October 20, 2020 @09:48PM
(#24844118)
Have a look at some published nn architectures - you'll typically see a few nonlinear units/transforms and a LOT of ReLU ones.