README.md

Neural network robustness verification via Integer Linear Programming

This is the code written in conjunction with the second part of my Master's thesis on GUM-compliant neural network robustness verification. The code was written for Python 3.10.

The final submission date is 23. January 2023. Until then, this code base will be subject to constant change.

Getting started

The INSTALL guide assists in installing the required packages.

Documentation

To locally build the HTML or pdf documentation first the required dependencies need to be installed into your virtual environment (check the INSTALL guide first and upon completion execute the following):

(venv) $ python -m piptools sync docs-requirements.txt
(venv) $ sphinx-build docs/ docs/_build
sphinx-build docs/ docs/_build
Running Sphinx v5.3.0
loading pickled environment... done
[...]
The HTML pages are in docs/_build.

After that the documentation can be viewed by opening the file docs/_build/index.html in any browser.

Roadmap

• implement first draft of optimization with dummy data to validate results