This is the code written in conjunction with the second part of the author's Master's thesis on GUM-compliant neural network robustness verification. The code was written for Python 3.10.
The final submission date was 23. January 2023.
The INSTALL guide assists in installing the required packages. After that you might want to have a look at our examples and/or the provided notebook to get a feeling for how to use the software.
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.
- check what improvements are made by switching to optimizable variables for the
$r_i$ s
This software is developed under the sole responsibility of Björn Ludwig (the author in the following). The software is made available "as is" free of cost. The author assumes no responsibility whatsoever for its use by other parties, and makes no guarantees, expressed or implied, about its quality, reliability, safety, suitability or any other characteristic. In no event will the author be liable for any direct, indirect or consequential damage arising in connection with the use of this software.
lp_nn_robustness_verification is distributed under the MIT license.