International Conference on Data Mining (ICDM), 2023 at Shanghai

Install the required packages:

python == 3.10
pip install -r requirements.txt

• CIFAR-10: the dataset can be download into the folder ./data/cifar10 automatically when running the main_cifar10 if the dataset is not existing.
• KMNIST: run ./data/kmnist/download.sh to download the dataset into folder ./data/kmnist.

For CIFAR10 dataset,

cd ./command
bash kmeans_select.sh

For KMNIST dataset,

cd ./command
bash kmeans_select_k.sh

This work uses code fragments from many projects. We acknowledge the authors of the following projects we referred in our codes.

• pykeops for kNN and k-Means.
• USL for parts of partitioned_kNN and get_selection_with_reg.