mnist and cifar10, nin not supported yet

.gitignore

@@ -0,0 +1,113 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+env/
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*,cover
+.hypothesis/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# pyenv
+.python-version
+
+# celery beat schedule file
+celerybeat-schedule
+
+# SageMath parsed files
+*.sage.py
+
+# dotenv
+.env
+
+# virtualenv
+.venv
+venv/
+ENV/
+
+# Spyder project settings
+.spyderproject
+
+# Rope project settings
+.ropeproject
+
+# data
+data/
+data
+
+# PyCharm
+.idea/
+
+# macOS
+.DS_Store
+
+# output directory
+output/
+
+# python import cache
+*.pyc
+
+# nfs temp file
+.nfs*

README.md

@@ -1,2 +1,38 @@
 # deepdefense.pytorch
-Code coming soon!
+Code for NeurIPS 2018 paper [Deep Defense: Training DNNs with Improved Adversarial Robustness](https://papers.nips.cc/paper/7324-deep-defense-training-dnns-with-improved-adversarial-robustness).
+
+Deep Defense is recipe to improve the robustness of DNNs to adversarial perturbations. We integrate an adversarial perturbation-based regularizer into the training objective, such that the obtained models learn to resist potential attacks in a principled way.
+
+## Environments
+* Python 3.5
+* PyTorch 0.4.1
+* glog 0.3.1
+
+## Datasets and Reference Models
+For fair comparison with DeepFool, we follow it to use [matconvnet](https://github.com/vlfeat/matconvnet/releases/tag/v1.0-beta24) to pre-process data and train reference models for MNIST and CIFAR-10.
+
+Please download processed datasets and reference models (including MNIST and CIFAR-10) at [download link](https://drive.google.com/open?id=15xoZ-LUbc9GZpTlxmCJmvL_DR2qYEu2J).
+
+## Usage
+To train a Deep Defense LeNet model using default parameters on MNIST:
+
+```
+python3 deepdefense.py --pretest --dataset mnist --arch LeNet
+```
+
+Argument ```--pretest``` indicates evaluating performance before fine-tuning, thus we can check the performance of reference model.
+
+Currently we've implemented ```MLP``` and ```LeNet``` for mnist, and ```ConvNet``` for CIFAR-10.
+
+## Citation
+Please cite our work in your publications if it helps your research:
+
+```
+@inproceedings{yan2018deep,
+  title={Deep Defense: Training DNNs with Improved Adversarial Robustness},
+  author={Yan, Ziang and Guo, Yiwen and Zhang, Changshui},
+  booktitle={Advances in Neural Information Processing Systems},
+  pages={417--426},
+  year={2018}
+}
+```

datasets/cifar10.py

@@ -0,0 +1,32 @@
+import torch.utils.data
+import numpy as np
+
+
+class CIFAR10Dataset(torch.utils.data.Dataset):
+    def __init__(self, phase='train', num_val=5000):
+        import scipy.io as sio
+        imdb = sio.loadmat('data/cifar10-data-ce5d97dd.mat')
+        images = imdb['images'][0][0][0].transpose()
+        sets = imdb['images'][0][0][2].flatten()
+        labels = (imdb['images'][0][0][1].flatten() - 1).astype(np.int64)
+        train_idx = np.where(sets == 1)[0][num_val:]
+        val_idx = np.where(sets == 1)[0][:num_val]
+        trainval_idx = np.where(sets == 1)[0]
+        test_idx = np.where(sets == 3)[0]
+        assert phase in ['train', 'val', 'trainval', 'test']
+        self.images = eval('images[%s_idx]' % phase)
+        self.labels = eval('labels[%s_idx]' % phase)
+        self.perm = np.arange(self.labels.size)
+
+    def __getitem__(self, index):
+        if np.random.rand() > 0.5:
+            images = np.fliplr(self.images[self.perm[index]]).copy()
+        else:
+            images = self.images[self.perm[index]]
+        return images, self.labels[self.perm[index]]
+
+    def __len__(self):
+        return self.labels.size
+
+    def shuffle(self, perm):
+        self.perm = perm

datasets/mnist.py

@@ -0,0 +1,33 @@
+import torch.utils.data
+import numpy as np
+import scipy.io as sio
+
+
+num_val = 10000  # first num_val examples in training set is used as validation set
+
+
+class MNISTDataset(torch.utils.data.Dataset):
+    def __init__(self, phase='train'):
+        imdb = sio.loadmat('data/mnist-data-0208ce21.mat')
+        images = imdb['images'][0][0][0].transpose()
+        sets = imdb['images'][0][0][3].flatten()
+        labels = imdb['images'][0][0][2].flatten() - 1
+        train_idx = np.where(sets == 1)[0][num_val:]
+        val_idx = np.where(sets == 1)[0][:num_val]
+        trainval_idx = np.where(sets == 1)[0]
+        test_idx = np.where(sets == 3)[0]
+        mean = imdb['images'][0][0][1].transpose()
+        assert phase in ['train', 'val', 'trainval', 'test']
+        self.images = eval('images[%s_idx]' % phase)
+        self.labels = eval('labels[%s_idx]' % phase)
+        self.mean = mean
+        self.perm = np.arange(self.labels.size)
+
+    def shuffle(self, perm):
+        self.perm = perm
+
+    def __getitem__(self, index):
+        return self.images[self.perm[index]], self.labels[self.perm[index]]
+
+    def __len__(self):
+        return self.labels.size