Probabilistic-Bilevel-Coreset-Selection

Requirements:

Pytorch 1.7
Python 3.7.7
CUDA Version 10.1
pyyaml 5.3.1
tensorboard 2.2.1
torchvision 0.5.0
tqdm 4.50.2

Command

Below are the commands for replicating the results of coreset selection and pixel selection experiments.

CUDA_VISIBLE_DEVICES=0  python cnn_mnist_probability_1step_pixel_shared_rein.py --inner_lr 5e-3 --inner_optim adam --test_freq 10 --outer_lr 1e-1 --max_outer_iter 2000 --limit 1000 --coreset_size 100 --div_tol 2 --epoch_converge 300 --runs_name outer1e-1_limit1000_size100_shared --wandb --vr --K 5 --clip_grad --test_freq 5 --start_coreset_size 784 --iterative

CUDA_VISIBLE_DEVICES=0  python cnn_mnist_probability_1step_reinforce.py --limit 1000  --inner_lr 5e-3 --inner_optim Adam --outer_lr 5e-2 --coreset_size 200 --runs_name size200 --wandb --iterative --project coreset_size_iterative  --epoch_converge 300 --test_freq 10 --max_outer_iter 2000 --wandb --vr --K 5 --clip_grad --test_freq 5

Cite

If you find this implementation is helpful to your work, please cite

@inproceedings{coreset,
  title={Probabilistic Bilevel Coreset Selection},
  author={Zhou, Xiao and Pi, Renjie and Zhang, Weizhong and Lin, Yong and Zhang, Tong},
  booktitle={International Conference on Machine Learning},
  year={2022},
  organization={PMLR}
}