Residual Connections Encourage Iterative Inference
Authors:
Preprint
Abstract
Residual networks (Resnets) have become a prominent architecture in deep learning. However, a comprehensive understanding of Resnets is still a topic of ongoing research. A recent view argues that Resnets perform iterative refinement of features. We attempt to further expose properties of this aspect. To this end, we study Resnets both analytically and empirically. We formalize the notion of iterative refinement in Resnets by showing that residual connections naturally encourage features of residual blocks to move along the negative gradient of loss as we go from one block to the next. In addition, our empirical analysis suggests that Resnets are able to perform both representation learning and iterative refinement. In general, a Resnet block tends to concentrate representation learning behavior in the first few layers while higher layers perform iterative refinement of features. Finally we observe that sharing residual layers naively leads to representation explosion and counterintuitively, overfitting, and we show that simple existing strategies can help alleviating this problem.Keywords
All Related Versions
This publication has 5 references indexed in Scilit:
- ImageNet classification with deep convolutional neural networksCommunications of the ACM, 2017
- The Time-Course of Ultrarapid Categorization: The Influence of Scene Congruency and Top-Down ProcessingI-Perception, 2016
- Identity Mappings in Deep Residual Networks2016
- Deep Residual Learning for Image Recognition2016
- Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification2015