xAI Dissertation Repository

This repository is the foundation for dissertation research in explainable AI, robustness, distribution shift, and stability.

Repository Structure

• data/ — raw and processed datasets for experiments
• models/ — trained model artifacts and serialized pipelines
• explainers/ — explanation methods, analysis scripts, and comparison code
• shifts/ — distribution shift scenarios, drift datasets, and shift generation utilities
• analysis/ — evaluation code, figures, statistical analysis, and notebooks
• results/ — experimental outputs, metrics, and reports
• src/ — project source code and reproducibility helpers
• docs/ — documentation for reproducibility, experiment tracking, and methodology

Setup Instructions

Python environment

1. Install Python 3.10 or newer.
2. Create a virtual environment:

python -m venv .venv
.\.venv\Scripts\Activate.ps1

3. Install pinned dependencies:

python -m pip install --upgrade pip
python -m pip install -r requirements.txt

Experiment tracking

This project uses MLflow for experiment tracking.

• Default tracking URI: ./mlruns
• Default experiment name: xai_dissertation

Run the MLflow UI locally:

mlflow ui --backend-store-uri file:./mlruns

Use src/mlflow_setup.py to configure experiments programmatically.

Reproducibility and seed discipline

Use src/config.py as the single source of truth for random seed and hashing salt.

from src.config import RANDOM_SEED, get_deterministic_hash

Record the commit hash for every experiment:

git rev-parse HEAD

License

This repository is licensed under the MIT License. See LICENSE for details.

Notes

• This repository captures the dissertation research pipeline and should remain reproducible from the first commit.
• Keep data artifacts out of Git by using the .gitignore rules and storing large datasets outside the repository if needed.