|
@ -12,14 +12,24 @@ A pytorch implementation for paper *[Unsupervised Domain Adaptation by Backpropa |
|
|
- Python 2.7/3.6 |
|
|
- Python 2.7/3.6 |
|
|
- PyTorch 0.3.1post2 |
|
|
- PyTorch 0.3.1post2 |
|
|
|
|
|
|
|
|
|
|
|
## Note |
|
|
|
|
|
|
|
|
|
|
|
- `Config()` 为针对特定任务的配置参数。 |
|
|
|
|
|
- `MNISTmodel()` 完全按照论文中的结构,但是 feature 部分添加了 `Dropout2d()`,实验发现是否添 |
|
|
|
|
|
加 `Dropout2d()` 对于最后的性能影响很大。最后实验重现结果高于论文,因为使用了额外的技巧,这里 |
|
|
|
|
|
还有值得探究的地方。 |
|
|
|
|
|
- `SVHNmodel()` 无法理解论文中提出的结构,为自定义结构。最后实验重现结果完美。 |
|
|
|
|
|
|
|
|
## Result |
|
|
## Result |
|
|
|
|
|
|
|
|
results of the default `params.py` |
|
|
|
|
|
|
|
|
| | MNIST-MNISTM | SVHN-MNIST | |
|
|
|
|
|
| :-------------: | :------------: | :--------: | |
|
|
|
|
|
| Source Only | 0.5225 | 0.5490 | |
|
|
|
|
|
| DANN | 0.7666 | 0.7385 | |
|
|
|
|
|
| This Repo | 0.8400 | 0.7339 | |
|
|
|
|
|
|
|
|
| | SVHN (Source) | MNIST (Target)| |
|
|
|
|
|
| :--------------------------------: | :------------: | :-----------: | |
|
|
|
|
|
| Source Classifier | 92.92% | 68.66% | |
|
|
|
|
|
| DANN | | ----% | |
|
|
|
|
|
|
|
|
- MNIST-MNISTM: `python mnist_mnistm.py` |
|
|
|
|
|
- SVHN-MNIST: `python svhn_mnist.py` |
|
|
|
|
|
|
|
|
## Other implementations |
|
|
## Other implementations |
|
|
|
|
|
|
|
@ -27,6 +37,8 @@ results of the default `params.py` |
|
|
- TensorFlow, <https://github.com/pumpikano/tf-dann> |
|
|
- TensorFlow, <https://github.com/pumpikano/tf-dann> |
|
|
- Theano, <https://github.com/shucunt/domain_adaptation> |
|
|
- Theano, <https://github.com/shucunt/domain_adaptation> |
|
|
- PyTorch, <https://github.com/fungtion/DANN> |
|
|
- PyTorch, <https://github.com/fungtion/DANN> |
|
|
|
|
|
- numpy, <https://github.com/GRAAL-Research/domain_adversarial_neural_network> |
|
|
|
|
|
- lua, <https://github.com/gmarceaucaron/dann> |
|
|
|
|
|
|
|
|
## Credit |
|
|
## Credit |
|
|
|
|
|
|
|
|