GitHub - rhaldarpurdue/Adversarial_Vulnerability_Convergence: Adversarial Vulnerability as a Consequence of Manifold Inseparibility

Command for ADAM clean training: python -u convergence.py --data ${D} --seed ${SEED} --epochs 1000

Command for clean training with batch normalization: python -u convergence_bn.py --data ${D} --seed ${SEED} --epochs 1000

Command for clean training with second-order optimization using KFAC: python -u mnist_kfac.py --data ${D} --seed ${SEED} --epochs 1000

Arguments: ${D}="MNIST" or "FashionMNIST" or "CIFAR10"

Name		Name	Last commit message	Last commit date
Latest commit History 6 Commits
.gitattributes		.gitattributes
cifar.zip		cifar.zip
convergence.py		convergence.py
convergence_bn.py		convergence_bn.py
data_utils.py		data_utils.py
mnist_kfac.py		mnist_kfac.py
readme.md		readme.md
utils.py		utils.py

Provide feedback