Back to Search
Start Over
Rethinking Skip Connection with Layer Normalization
- Source :
- COLING
- Publication Year :
- 2020
- Publisher :
- International Committee on Computational Linguistics, 2020.
-
Abstract
- Skip connection is a widely-used technique to improve the performance and the convergence of deep neural networks, which is believed to relieve the difficulty in optimization due to non-linearity by propagating a linear component through the neural network layers. However, from another point of view, it can also be seen as a modulating mechanism between the input and the output, with the input scaled by a pre-defined value one. In this work, we investigate how the scale factors in the effectiveness of the skip connection and reveal that a trivial adjustment of the scale will lead to spurious gradient exploding or vanishing in line with the deepness of the models, which could by addressed by normalization, in particular, layer normalization, which induces consistent improvements over the plain skip connection. Inspired by the findings, we further propose to adaptively adjust the scale of the input by recursively applying skip connection with layer normalization, which promotes the performance substantially and generalizes well across diverse tasks including both machine translation and image classification datasets.
- Subjects :
- Linear component
Normalization (statistics)
Machine translation
Contextual image classification
Artificial neural network
Computer science
05 social sciences
010501 environmental sciences
computer.software_genre
01 natural sciences
0502 economics and business
050207 economics
Spurious relationship
computer
Algorithm
0105 earth and related environmental sciences
Subjects
Details
- Database :
- OpenAIRE
- Journal :
- Proceedings of the 28th International Conference on Computational Linguistics
- Accession number :
- edsair.doi...........08d0c01f5509a687bd25e846913f076a