aminul-huq / mixPGD

mixPGD

This is the implementation for the mixPGD adversarial training paper.

Training type

File name	Training Type
standard_main.py	Traditional training
FGSM_main.py	FGSM adversarial training
PGD_main.py	PGD adversarial training
feature_scatter_main.py	Feature Scattering adversarial training
mixPGD_main.py	mixPGD adversarial training

How to run files

just use the following command on any of the file name mentioned above

python filename.py

dataset

We have used LibriSpeech dataset in this experiment. You need to define the location of the dataset inside the .py files.

Hyper-parameters

All the hyper-parameters can be changed in the individual .py files.

If any problems are faced regarding the code please contact at: [e-mail]