DeepSplit is a library for solving neural network verification problems through the Alternative Direction Method of Multipliers (ADMM). Details of this method can be found in this paper:
- DeepSplit: Scalable Verification of Deep Neural Networks via Operator Splitting
Shaoru Chen*, Eric Wong*, J. Zico Kolter, Mahyar Fazlyab (* Equal contribution)
Under review of IEEE Open Journal of Control Systems
Run the following commands to install a conda environment:
conda create -n deepsplit python=3.10
conda activate deepsplit
# Example: install pytorch on macOS.
conda install pytorch==1.12.1 torchvision==0.13.1 torchaudio==0.12.1 -c pytorch
pip install auto-lirpa
conda install -c conda-forge cvxpy
pip install convex_adversarial
pip install -r requirements.txt
Run the following command to bound the output of an MLP image classifier on the MNIST dataset:
python examples/mnist/main_mnist_fc.py