This repository contains code and experiment data related to the work titled Exploring Classifiers with Differentiable Decision Boundary Maps, due to A. Machado, M. Behrisch, and A. Telea, to appear in the proceedings of EuroVis 2024.
In this project, I am exploring how to augment Decision Boundary Maps (DBMs) with information coming from Adversarial Example Generation (AEG) theory.
The AEG method implemented so far is DeepFool: Moosavi-Dezfooli, Seyed-Mohsen, Alhussein Fawzi, and Pascal Frossard. “DeepFool: A Simple and Accurate Method to Fool Deep Neural Networks.” arXiv, July 4, 2016. https://doi.org/10.48550/arXiv.1511.04599.
I also propose additional views based off of differentiating specific components in the DBM-generating process.
You will need:
- Poetry installed
- Python >=3.9
- TkInter and its Python bindings. Something like
sudo apt install python3-tk
.
From the directory containing pyproject.toml
.
poetry install
poetry run main