Adaout

Adaout is a practical and flexible regularization method with high generalization and interpretability.

Requirements

• python 3.6 (Anaconda version >=5.2.0 is recommended)
• torch (torch version >=1.1.0 is recommended)
• torchvision (torchvision version >=0.3.0 is recommended)
• pandas
• numpy
• NVIDIA GPU + CUDA CuDNN

Datasets

• CIFAR-10, CIFAR-100, SVHN, ImageNet and others

Getting started

• Download datasets and extract it inside data
• Train: python train.py, python train100.py or python train_svhn.py
• Evaluate:
  • Pretrained models for CIFAR-10 and CIFAR-100 are available at this link. Download and extract them in the save_model/resnet56_10 or save_model/resnet56_100 directory.
  • You should achieve about 94.63% accuracy on CIFAR-10, and 74.18% accuracy on CIFAR-100 datasets.