AdversarialNetworks

Using adversarial networks to reduce uncertainties or look for BSM physics

Tests and Tutorials

You will need to run this in the srappoccio/ubccr-cms:latest docker image.

(Optional) Log into winterfell (every time)

Open a terminal and ssh to winterfell:

ssh -i ~/(yourpemfile).pem (yourusername)@199.109.192.91

If you've set up your account correctly, if you type pwd you should see /mnt/users/(yourusername).

Set up a working directory (do once)

Make a directory for your work. Make sure it has write access:

mkdir dockers
chmod 777 dockers

Clone the package:

git clone https://github.com/ubcms-xai/AdversarialNetworks.git
chmod -R 777 AdversarialNetworks

Make a data directory:

cd AdversarialNetworks
mkdir data
chmod -R 777 .

Run the docker image and jupyter notebook (do every time)

Start the docker image. You may have to change the port from 8888 to something else like 8883, 8884, etc.

bash ./runDockerCommandLine.sh 8888 srappoccio/ubccr-cms:latest

Start a jupyter notebook:

jupyter notebook --ip 0.0.0.0 --no-browser

Run the code

Then you can begin running the code.

1. Open ToyModel/makeFourVectors.ipynb
2. Click "Run"
3. Change addPerturbation = True to addPerturbation = False
4. Click "Run" again
5. Open ToyModel/toy_adversarial_1dcnn.ipynb
6. Click "Run"