pytorch-VAT
pytorch-VAT copied to clipboard
Unofficial pytorch implementation of a paper, Distributional Smoothing with Virtual Adversarial Training [Miyato+, ICLR2016].
pytorch-VAT
This is an unofficial pytorch implementation of a paper, Distributional Smoothing with Virtual Adversarial Training [Miyato+, ICLR2016].
Please note that this is an ongoing project.
Requirements
- Python 3.5+
- PyTorch 0.4
- TorchVision
- click
Usage
Train MNIST classifier with only labeled data (100 images)
CUDA_VISIBLE_DEVICES=<gpu_id> python train_baseline.py --n_label 100
Error rate: about 30%
VAT with mixture of labeled and unlabeled data
CUDA_VISIBLE_DEVICES=<gpu_id> python train_vat.py --n_label 100
Error rate: about 2%
References
- [1]: T. Miyato et al. "Distributional Smoothing with Virtual Adversarial Training", in ICLR, 2016.