Robust width adversarial defense

This repository contains the code for our paper, Robust width adversarial defense.

Setup

All dependencies are described in the environment.yml file. You can setup this environment using Anaconda:

conda env create -f environment.yml
conda activate sparse-ds

Hyperparameter optimization

The hyperparameter optimization is carried out by the main.py Python script. For our experiments, we saved the results of the Optuna trials to a local SQLite database called fourier.db. The results can be viewed using the Optuna dashboard:

optuna-dashboard sqlite:///fourier.db

Reproducing experiments

You can reproduce our experiments by running the appropriate script:

bash experiments.sh

This will run all of the experiments one by one. Individual adversarial attacks can be carried out using the attack.py Python script.