my code is first inspired by

https://github.com/CuthbertCai/pytorch_DANN

Domain-Adversarial Training of Neural Networks

https://arxiv.org/pdf/1505.07818.pdf
implement Domain-Adversarial Training of Neural Networks by pytorch

Dataset

• mnist
• mnist_m
• download from https://github.com/fungtion/DANN

result

	original paper	my implementation
Source only	0.52	0.5
DANN	0.766	0.82~0.83

TODO

• try to tune toy example better
• use less parameters to train the model