Code for our paper
1-Lipschitz Neural Networks are more expressive with N-Activations.
It includes code for the proposed
- Python 3.9
- PyTorch 1.12
Use the settings_nr
(see below or settings file) to specify the setting and run:
python train.py 10
The last digit of the settings number determines the activation, 0 means MaxMin and 1 for the
Settings Nr | Dataset | Method |
---|---|---|
10-19 | CIFAR-10 | AOL |
20-29 | CIFAR-10 | CPL |
30-39 | CIFAR-10 | SOC |
40-49 | CIFAR-100 | AOL |
50-59 | CIFAR-100 | CPL |
60-69 | CIFAR-100 | SOC |
70-79 | Tiny ImageNet | AOL |
80-89 | Tiny ImageNet | CPL |
90-99 | Tiny ImageNet | SOC |