The implementation of "Enhancing Mixup-based Semi-Supervised Learningwith Explicit Lipschitz Regularization" [ICDM 2020].
python train.py
(for baseline Mixup)
python train.py --ALR
(for the proposed Mixup-LR)
- PyTorch
- torchvision
(There might be more requirements but shouldn't be difficult to install them using conda.)
In order to use all data, a separate class CIFAR10All
is created inside cifar.py of torchvision. The only difference of this class than the regular CIFAR10
is that it's train list also comprises of test_batch
beside reguarl data_batch_i
.