Skip to content

columbia/pixeldp

master
Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Research paper: https://arxiv.org/abs/1802.03471

Train the model.

Chose the model ll.174-176, and the parameters ll. 111-142 in main.py. See models/params.py for parameter usage. Then, for instance:

python3 main.py --num_gpus 1 --dataset cifar10 --mode train

Eval the model.

Chose parameters in main.py, then:

python3 main.py --num_gpus 1 --dataset cifar10 --mode eval

The eval data is logged in eval_data.json

Other "modes"

attack, attack_eval, plot

For the attack, the attack type and parameters are ll. 152-181 in main.py.

Files:

The most basic model is in models/pixeldp_cnn.py, with robustness tests in models/utils/robustness.py, and trained/evaluated through models/{train,evaluate}.py.

Attacks are implemented in attacks/carlini.py and attacks/pgd.py.

Pretrained model:

The ResNet for cifar10, L=0.1 (for L2 attacks) is available here: http://www.cs.columbia.edu/~mathias/pixeldp/cifar10.zip

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages