ShakeDrop Regularization for Deep Residual Learning
PyramidNet + Shake Drop + Cosine Annealing Learning Rate in Chainer with Colaboratory (without Random Erasing)
Although I made PyramidNet + Shake Drop implimentation code, in Colaboratory it takes 15 days in training (LoL).
Therefore, I can't train this network.
If you want to train this network, please change format (i.e. ipynb -> py) after downloading this code.
But, since I didn't finish to train this network, I can't ensure whether or not this code is collect.
Great thanks to @shunk031! (PyramidNet in Chainer)
・ Chainer 5.0.0 (Google Colaboratory at 25/3/2019)