02/02/2021

Delving into Variance Transmission and Normalization: Shift of Average Gradient Makes the Network Collapse

Yuxiang Liu, Jidong Ge, Chuanyi Li, Jie Gui

Keywords:

Abstract Paper Similar Papers

Abstract: Normalization operations are essential for state-of-the-art neural networks and enable us to train a network from scratch with a large learning rate (LR). We attempt to explain the real effect of Batch Normalization (BN) from the perspective of variance transmission by investigating the relationship between BN and Weights Normalization (WN). In this work, we demonstrate that the problem of the shift of the average gradient will amplify the variance of every convolutional (conv) layer. We propose Parametric Weights Standardization (PWS), a fast and robust to mini-batch size module used for conv filters, to solve the shift of the average gradient. PWS can provide the speed-up of BN. Besides, it has less computation and does not change the output of a conv layer. PWS enables the network to converge fast without normalizing the outputs. This result enhances the persuasiveness of the shift of the average gradient and explains why BN works from the perspective of variance transmission. The code and appendix will be made available on https://github.com/lyxzzz/PWSConv.

The video of this talk cannot be embedded. You can watch it here:
https://slideslive.com/38947810
(Link will open in new window)
 0
 0
 0
 0
  Share    
This is an embedded video. Talk and the respective paper are published at AAAI 2021 virtual conference. If you are one of the authors of the paper and want to manage your upload, see the question "My papertalk has been externally embedded..." in the FAQ section.

Comments

Post Comment
no comments yet
code of conduct: tbd Characters remaining: 140

Similar Papers

 0
 0
 0
 0
 1:01 
 0
 0
 0
 0
 4:08 
 0
 0
 0
 0
 5:16 
 0
 0
 0
 0
 5:00 
 0
 0
 0
 0
 3:27 
 0
 0
 0
 0
 0:59 
 0
 0
 0
 0
 12:51 
 0
 0
 0
 0
 3:23 
 0
 0
 0
 0
 4:58 
 0
 0
 0
 0
 4:57 
 0
 0
 0
 0
 1:00 
 0
 0
 0
 0
 5:17 
 0
 0
 0
 0
 16:12 
 0
 0
 0
 0
 1:00 
 0
 0
 0
 0
 1:01 
 0
 0
 0
 0
 5:30 
 0
 0
 0
 0
 6:45 
 0
 0
 0
 0
 14:55 
 0
 0
 0
 0
 3:05 
 0
 0
 0
 0
 15:00 
 0
 0
 0
 0
 14:41 
 0
 0
 0
 0
 10:57 
 0
 0
 0
 0
 14:35 
 0
 0
 0
 0
 4:36 
 0
 0
 0
 0
 3:05 
 0
 0
 0
 0
 12:49 
 0
 0
 0
 0
 5:00 
 0
 0
 0
 0
 12:58 
 0
 0
 0
 0
 1:01 
 0
 0
 0
 0
 5:22 
 0
 0
 0
 0
 6:25 
 0
 0
 0
 0
 17:13 
 0
 0
 0
 0
 5:07 
 0
 0
 0
 0
 14:35 
 0
 0
 0
 0
 12:26 
 0
 0
 0
 0
 3:23 
 0
 0
 0
 0
 5:06 
 0
 0
 0
 0
 5:02 
 1
 0
 0
 0
 14:15 
 0
 0
 0
 0
 12:43