What's this

Implementation of Weighted Residuals for Very Deep Networks Depth (WResNet) by chainer

Dependencies

git clone https://github.com/nutszebra/weighted_residual_net.git
cd weighted_residual_net
git submodule init
git submodule update

How to run

python main.py -g 0

Details about my implementation

All hyperparameters and network architecture are the same as in [1] except for some parts.

• Data augmentation
  Train: Pictures are randomly resized in the range of [32, 36], then 32x32 patches are extracted randomly and are normalized locally. Horizontal flipping is applied with 0.5 probability.
  Test: Pictures are resized to 32x32, then they are normalized locally. Single image test is used to calculate total accuracy.

• Drop probability
  The linear decay is used. P_0 is 1 and P_L is 0.5.

Cifar10 result

network	depth	total accuracy (%)
WResNet-d [1]	1192	95.3
my implementation	1192	soon

References

Weighted Residuals for Very Deep Networks [1]