Given a composed of N layers, usually some layers are more “important” than other layers (eg. removing i hurts performance more than removing j). Is there any theory that can be used to the effectiveness of each neural layer?



Source link

LEAVE A REPLY

Please enter your comment!
Please enter your name here