Code for co-training ResNet50 backbone model on human EEG to improve adversarial robustness

• torch
• torchvision
• numpy
• matplotlib
• scipy
• foolbox 3.x

-DataPreparition provides .ipynb file to create EEG-image pairs for training and validating, shuffled biological EEG and random EEG for ablation study.

• Model_training provides scripts for EEG co-training.
• Robustness_evaluation contains scripts for adversarial robustness evaluation.
• Correlation_analysis contains scripts for measuring correlation between EEG prediction accracy and adversarial robustness gain.

• (A.T. Gifford, et.al) A large and rich EEG dataset for modeling human visual object recognition