Skip to content

This repo contains the implementation of ASK attack/defense proposed by us

Notifications You must be signed in to change notification settings

tqch/adversarial-soft-knn

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

16 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

ASK: Adversarial Soft KNN [paper]

CIFAR10 ASK robust training

python ask_train.py --dataset=cifar10

Imagenette ASK robust training

python ask_train.py --dataset=imagenette --batch-size=64 --lr=0.01 --eps-train=2 --eps-eval=4 --eps-ask=4 --step-size=1 --dknn-size=8500 --hidden-layer=2

Note 1: please add d flag the first time you run a model on a dataset

Note 2: please use the disable-ask option to run standard adversarial training

About

This repo contains the implementation of ASK attack/defense proposed by us

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages