xai-lab-ese

The 10 XAI lab experiments as a pip package. Each experiment has:

• theory — how to solve that type of problem: exam task, formulas, step-by-step method, worked example with numbers, notes on common mistakes, viva questions;
• exam code: short programs to solve the numericals (plain Python, no libraries);
• lab code: the original lab notebook code, plus the .ipynb itself.

Install

pip install --upgrade xai-lab-ese            # theory + exam code (no dependencies)
pip install --upgrade "xai-lab-ese[lab]"     # + numpy, pandas, matplotlib, scikit-learn for the lab notebooks
pip install --upgrade "xai-lab-ese[all]"     # + tensorflow (lab Exp 9)

Colab / Jupyter: !pip install --upgrade xai-lab-ese

Command line

xai-lab-ese --list                     # list the 10 experiments
xai-lab-ese --show lime                # title + theory + exam code
xai-lab-ese --show lime --lab          # ... with the lab-notebook code
xai-lab-ese --show lime --both         # ... with both
xai-lab-ese --theory shapley           # only the theory / how to solve
xai-lab-ese --exam gradcam             # only the exam code
xai-lab-ese --code gradcam             # only the lab-notebook code
xai-lab-ese --save lrp --out lrp.py    # save exam code (add --lab for notebook code)
xai-lab-ese --notebook pdp             # save the original Exp4.ipynb
xai-lab-ese --save-all --out xai_code  # everything: exam/lab .py + .txt, theory, notebooks
xai-lab-ese --run xrai                 # run the exam code (add --lab for the notebook code)

Python

from xai_lab_ese import list_experiments, show_experiment, get_theory, get_exam_code, get_lab_code, save_all

list_experiments()
show_experiment("lime")                 # title, theory, then the exam code
show_experiment("lime", code="lab")     # lab-notebook code instead
print(get_theory("shapley"))
code = get_exam_code("regression")
save_all("xai_code")

Experiments

Experiments can also be selected by number (1–10) or exp1–exp10.

See PUBLISHING.md for how to upload to PyPI.

License

MIT