Note:

We provide the implementation of our methods in this codebase. 

main.py: includes the implementation of DSA and DSA + Ours
main_DM.py: includes the implementation of DM and DM + Ours
utils.py: includes the code for loading datasets and some common shared functions
networks.py: includes the model architecture we use in the paper.

The code can be run using the same way as running a normal DM or DSA algorithm plus the option --use_kde. No other hyperparameters are required. We also need the embeddings generated from SelecMix. The steps can be found here: https://github.com/bluemoon010/selecmix

The official codebase will be released on github in the future. Thank you!!!