This repository contains code for the paper:
Verifying Binarized Neural Networks by Angluin-Style Learning
"Verifying Binarized Neural Networks by Angluin-Style Learning"
Andy Shih, Adnan Darwiche, Arthur Choi
In 22nd International Conference on Theory and Applications of Satisfiability Testing (SAT), 2019
@inproceedings{SDCsat19,
author = {Andy Shih and Adnan Darwiche and Arthur Choi},
title = {Verifying Binarized Neural Networks by Angluin-Style Learning},
booktitle = {Proceedings of the 22nd International Conference on Theory and Applications of Satisfiability Testing (SAT)},
month = {july},
year = {2019},
keywords = {conference}
}
Requires: Python2 torch Java cmake3
First you have to build the riss-solver coprocessor Follow the setups in https://github.com/nmanthey/riss-solver/blob/master/doc/TUTORIAL.md In particular:
cd bnn/BinaryNet/cpp/
git clone https://github.com/nmanthey/riss-solver.git
cd riss-solver
mkdir -p build
cd build
cmake ..
make coprocessor
Use the ./run_start script to start the experiments.
Change the arguments [num_features, num_hidden, radius, type] as needed.
Check out the paper for more details.
For questions, contact us at:
andyshih at cs dot stanford dot edu