DynamicFeatureAggregation

This is an official pytorch implementation of the paper "Robust Representation via Dynamic Feature Aggregation"

Requirements

• numpy>=1.19.2
• Pillow>=8.3.1
• pytorch>=1.2.0
• torchvision>=0.4.0
• tqdm>=4.62.3

Datasets

• CIFAR-10/100 datasets can be downloaded by torchvision. Note that you can set torchvision.datasets.CIFAR10/100(..., download=True) in ./train.py to download the corresponding dataset and keep the directory path.

Usage

• Train Wide_ResNet-28-10 on CIFAR10 with GPU 0, ./data/ is your directory path of dataset.

bash run.sh

Then the model is saved at ./model/cifar10/, where TOP_1_Net.pth refers to the model with best clean accuracy.

• or directly give the configuration by

CUDA_VISIBLE_DEVICES=${gpu_device} \
python train.py \
    --lr 0.1 \
    --depth ${depth} \
    --widen_factor ${width} \
    --dataset ${dataset} \
    --savedir ${savedir} \
    --alpha ${predefined param. for Beta distribution} \
    --noise ${noise term for aggregation} \

Related Repository

• The adopted Wide_ResNet is based on https://github.com/meliketoy/wide-resnet.pytorch

• Our adversarial attacks are conducted through https://github.com/Harry24k/adversarial-attacks-pytorch

• The protocol and basic OOD test method can be found at https://github.com/facebookresearch/odin and https://github.com/zaeemzadeh/OOD, respectively.