ATEX_code
Please run the src/experiment2.py to reproduce the experiment results.
-
train.py
function train(): Train baseline model -
advtrain_explain_tangval.py
function advTrain_explanation_indirect(): Train ATEX model -
advtrain_explain_tangval.py
function advTrain_FGSM(): Evaluate model adversairal defence performance -
attack_explain.py
function attack_explanation(): Evaluate model robustness before and after defense -
compare_explain.py
function compare_explanation(): Compare model attention maps with SmoothGrad before and after defense -
attack_explain.py
function attack_explanation_random(): Visualization of Adversarial attack on interpretation -
attack_explain.py
function explanation_random(): Visualization saliency map of ATEX model and Baseline model
we have clear code annotations to help you to understand the code.