Deep Neural Network Analysis SuSe2024

Blackbox Interpretability - LIME

This repository is from the 2024 course "Deep Neural Network Analysis" from the university of Osnabrück, held by Lukas Niehaus. Topic for this group project is methods to interpret blackbox models using LIME. A presentation PDF and scripts with visualizations are provided here, the blackbox models used in this repository are taken from the repository for group 1 of the course.

Installation

1. Clone the repository:

   git clone https://github.com/madammann/DNNA24_blackbox_lime.git

2. Navigate to the project directory:

   cd DNNA-Blackbox-Interpretability---LIME

3. Install dependencies:

   • Using environment.yml

     conda env create -f envirnoment.yml

   • Using requirements.txt

     pip install -r requirements.txt

4. Download the model checkpoints folder from github.com/lucasld/neural_network_analysis.

5. Put the folder in the root directory of this repository.

About

Repository structure

This repository contains two jupyter notebooks, one called segmentation and one called lime. The Lime notebook contains the pipelines for image and tabular explanations used, the segmentation one visualizes how slic segmentation works.

Contact

Marlon Dammann mdammann@uni-osnabrueck.de
Iheb Marouaniimarouani@uni-osnabrueck.de

References

https://github.com/marcotcr/lime