Overview

This repository provides the implementation for the paper "Improved Regularization and Robustness for Fine-tuning in Neural Networks", which will be presented as a poster paper in NeurIPS'21.

In this work, we propose a regularized self-labeling approach that combines regularization and self-training methods for improving the generalization and robustness properties of fine-tuning. Our approach includes two components:

• First, we encode layer-wise regularization to penalize the model weights at different layers of the neural net.
• Second, we add self-labeling that relabels data points based on current neural net's belief and reweights data points whose confidence is low.

Requirements

To install requirements:

pip install -r requirements.txt

Data Preparation

We use seven image datasets in our paper. We list the link for downloading these datasets and describe how to prepare data to run our code below.

• Aircrafts: download and extract into ./data/aircrafts
  • remove the class 257.clutter out of the data directory
• CUB-200-2011: download and extract into ./data/CUB_200_2011/
• Caltech-256: download and extract into ./data/caltech256/
• Stanford-Cars: download and extract into ./data/StanfordCars/
• Stanford-Dogs: download and extract into ./data/StanfordDogs/
• Flowers: download and extract into ./data/flowers/
• MIT-Indoor: download and extract into ./data/Indoor/

Our code automatically handles the split of the datasets.

Usage

Our algorithm (RegSL) interpolates between layer-wise regularization and self-labeling. Run the following commands for conducting experiments in this paper.

Fine-tuning ResNet-101 on image classification tasks.

python train_constraint.py --model ResNet101 \
    --config configs/config_constraint_indoor.json \
    --reg_method constraint --reg_norm frob \
    --reg_extractor 0.136809975858091 --reg_predictor 6.40780158171339 --scale_factor 2.52883770643206\
    --device 1

python train_constraint.py --model ResNet101 \
    --config configs/config_constraint_aircrafts.json \
    --reg_method constraint --reg_norm frob \
    --reg_extractor 1.18330556653284 --reg_predictor 5.27713618808711 --scale_factor 1.27679969876201\
    --device 1

python train_constraint.py --model ResNet101 \
    --config configs/config_constraint_birds.json \
    --reg_method constraint --reg_norm frob \
    --reg_extractor 0.204403908747731 --reg_predictor 23.7850606577679 --scale_factor 4.73803591794678\
    --device 1

python train_constraint.py --model ResNet101 \
    --config configs/config_constraint_caltech.json \
    --reg_method constraint --reg_norm frob \
    --reg_extractor 0.0867998872549272 --reg_predictor 9.4552942790218 --scale_factor 1.1785989596144\
    --device 1

python train_constraint.py --model ResNet101 \
    --config configs/config_constraint_cars.json \
    --reg_method constraint --reg_norm frob \
    --reg_extractor 1.3340347414257 --reg_predictor 8.26940794089601 --scale_factor 3.47676759842434\
    --device 1

python train_constraint.py --model ResNet101 \
    --config configs/config_constraint_dogs.json \
    --reg_method constraint --reg_norm frob \
    --reg_extractor 0.0561320847651626 --reg_predictor 4.46281825974388 --scale_factor 1.58722606909531\
    --device 1

python train_constraint.py --model ResNet101 \
    --config configs/config_constraint_flower.json \
    --reg_method constraint --reg_norm frob \
    --reg_extractor 0.131991042311165 --reg_predictor 10.7674132173309 --scale_factor 4.98010215976503\
    --device 1

Fine-tuning ResNet-18 under label noise.

python train_label_noise.py --config configs/config_constraint_indoor.json --model ResNet18 \
    --reg_method constraint --reg_norm frob \
    --reg_extractor 7.80246991703043 --reg_predictor 14.077402847906 \
    --noise_rate 0.2 --train_correct_label --reweight_epoch 5 --reweight_temp 2.0 --correct_epoch 10 --correct_thres 0.9 

python train_label_noise.py --config configs/config_constraint_indoor.json --model ResNet18 \
    --reg_method constraint --reg_norm frob \
    --reg_extractor 8.47139398080791 --reg_predictor 19.0191127114923 \
    --noise_rate 0.4 --train_correct_label --reweight_epoch 5 --reweight_temp 2.0 --correct_epoch 10 --correct_thres 0.9 

python train_label_noise.py --config configs/config_constraint_indoor.json --model ResNet18 \
    --reg_method constraint --reg_norm frob \
    --reg_extractor 10.7576018531961 --reg_predictor 19.8157649727473 \
    --noise_rate 0.6 --train_correct_label --reweight_epoch 5 --reweight_temp 2.0 --correct_epoch 10 --correct_thres 0.9 
    
python train_label_noise.py --config configs/config_constraint_indoor.json --model ResNet18 \
    --reg_method constraint --reg_norm frob \
    --reg_extractor 9.2031662757248 --reg_predictor 6.41568500472423 \
    --noise_rate 0.8 --train_correct_label --reweight_epoch 5 --reweight_temp 1.5 --correct_epoch 10 --correct_thres 0.9 

Fine-tuning Vision Transformer on noisy labels.

python train_label_noise.py --config configs/config_constraint_indoor.json \
    --model VisionTransformer --is_vit --img_size 224 --vit_type ViT-B_16 --vit_pretrained_dir pretrained/imagenet21k_ViT-B_16.npz \
    --reg_method none --reg_norm none \
    --lr 0.0001 --device 1 --noise_rate 0.4

python train_label_noise.py --config configs/config_constraint_indoor.json \
    --model VisionTransformer --is_vit --img_size 224 --vit_type ViT-B_16 --vit_pretrained_dir pretrained/imagenet21k_ViT-B_16.npz \
    --reg_method none --reg_norm none \
    --lr 0.0001 --device 1 --noise_rate 0.8

python train_label_noise.py --config configs/config_constraint_indoor.json \
    --model VisionTransformer --is_vit --img_size 224 --vit_type ViT-B_16 --vit_pretrained_dir pretrained/imagenet21k_ViT-B_16.npz \
    --reg_method constraint --reg_norm frob \
    --reg_extractor 0.7488074175044196 --reg_predictor 9.842955837419588 \
    --train_correct_label --reweight_epoch 24 --correct_epoch 18\
    --lr 0.0001 --device 1 --noise_rate 0.4

python train_label_noise.py --config configs/config_constraint_indoor.json \
    --model VisionTransformer --is_vit --img_size 224 --vit_type ViT-B_16 --vit_pretrained_dir pretrained/imagenet21k_ViT-B_16.npz \
    --reg_method constraint --reg_norm frob \
    --reg_extractor 0.1568903647089986 --reg_predictor 1.407080880079702 \
    --train_correct_label --reweight_epoch 18 --correct_epoch 2\
    --lr 0.0001 --device 1 --noise_rate 0.8

Please follow the instructions in ViT-pytorch to download the pre-trained models.

Fine-tuning ResNet-18 on ChestX-ray14 data set.

Run experiments on ChestX-ray14 in reproduce-chexnet path:

cd reproduce-chexnet

python retrain.py --reg_method None --reg_norm None --device 0

python retrain.py --reg_method constraint --reg_norm frob \
    --reg_extractor 5.728564437344309 --reg_predictor 2.5669480884876905 --scale_factor 1.0340072757925474 \
    --device 0

Citation

If you find this repository useful or happen to use it in a research paper, please cite our work with the following bib information.

@article{li2021improved,
  title={Improved Regularization and Robustness for Fine-tuning in Neural Networks},
  author={Li, Dongyue and Zhang, Hongyang},
  journal={Advances in Neural Information Processing Systems},
  volume={34},
  year={2021}
}

Acknowledgment

Thanks to the authors of the following repositories for providing their implementation publicly available.

• mars-finetuning

• WS-DAN.PyTorch

• ViT-pytorch

• reproduce-chexnet