This is the official codebase of our NeurIPS 2020 paper "Robust Optimal Transport with Applications in Generative Modeling and Domain Adaptation".
Go to GAN folder.
To train robust WGAN experiments on CIFAR-10 dataset corrupted with MNIST, run
python main.py --base_cfg_path configs/experiments/CIFAR10_MNIST.json --cfg_path configs/unconditional/WGAN.json
By changing the configs and experiment in base configs, different models and datasets can be run.
To train domain adaptation models, go to DA folder and run
python main.py --cfg-path configs/robust_adversarial.json