SAND: One-Shot Feature Selection with Additive Noise Distortion

Code for SAND Paper - (ICML 2025). SAND is a feature selection algorithm for neural networks based on simple addition of a noise layer.

Experiments

This project was built using Python 3.9.2

To download the datasets run sh sand/experiments/get_all_data.sh

To run the experiments for SAND on the 6 standard datasets, run python -m sand.experiments.experiment

Note: To run the experiments for SAND on the other datasets, change the dataset name and the corresponding number of features to select in the experiment file. Additionally, you can change the algorithm to run the experiments for other methods.

Acknowledgements

This repository is an edited clone of Sequential Attention. The original code has been modified to suit the specific needs of this project. All credit for the original code goes to the authors of the original repository.