/Stochastic-Cardinality-in-Residual-Network

Primary LanguagePython

Stochastic-Cardinality-in-Residual-Network

This work proposes a new regularization method in ResNeXt from the cardinality dimension which is called Stochastic Cardinality. It's very similar to Stochastic Depth, another regularization method based on Residual network. The only difference is that Stochastic Depth deals with the depth dimension but Stochastic Cardinality deals with the cardinality dimension.

To complete this work and compare with other competitive results from different models, I also reproduce some related models including ResNets, ResNeXts and Stochastic Depth.

I use following papers and codes as reference:

  1. ResNet: Deep Residual Learning for Image Recognition | codes
  2. ResNeXt: Aggregated Residual Transformations for Deep Neural Networks | codes
  3. Stochastic Depth: Deep Networks with Stochastic Depth | codes