On the Noisy Gradient Descent that Generalizes as SGD
A PyTorch implementation for the experiments in the paper:
On the Noisy Gradient Descent that Generalizes as SGD
Jingfeng Wu, Wenqing Hu, Haoyi Xiong, Jun Huan, Vladimir Braverman, Zhanxing Zhu
Usage
FashionMNIST
See folder FashionMNIST
SVHN
See folder SVHN
CIFAR-10
See folder CIFAR-10
CIFAR-100
See folder CIFAR-100