constrained-nn-training

Constrained Neural Network Training via Reachability Analysis

Paper: https://arxiv.org/abs/2107.07696

Setup

(Developed and tested on Windows with GPU support)

1. Make sure anaconda is setup. Open anaconda prompt
2. Create conda environment: conda env create -f environment.yml
3. Activate newly created environment: conda activate reach-net
4. Install cvxpy: pip install cvxpy
5. Install cvxpylayers: pip install cvxpylayers
6. Install pypoman: pip install pypoman
7. Install (latest) pytorch with CUDA support: conda install pytorch torchvision torchaudio cudatoolkit=10.2 -c pytorch

Jupyter notebooks in VSCode

1. Open new anaconda prompt and activate environment
2. Run code to open VSCode
3. In bottom left corner of VSCode, select interpreter 'reach-net': conda
4. In upper right corner of VSCode, select kernel 'reach-net': conda (May need to go to Preferences > Search "path" and select "python" and add path to anaconda envs to "Venv Path")

Usage

Main notebook for generating results is constrained_opt.ipynb