MUGA LAB Undergraduate Research – Analysis Notebooks

Model Understanding and Generative Alignment Laboratory (MUGA LAB)
Department of Mathematics • Ateneo de Manila University BS Applied Mathematics (Data Science Track)

Overview

This folder contains Jupyter notebooks that provide visual, metric-based, and reproducibility analyses for undergraduate research projects in Calibration, Distillation, and Interpretability.

Each notebook complements a Python experiment script from ../experiments/ and is designed to be publication-ready, MLflow-tracked, and pedagogically structured.

Notebook Collection

Notebook	Purpose	Key Dependencies
`calibration_analysis.ipynb`	Computes and visualizes calibration metrics (ECE, MCS, NLL). Compares pre- and post-temperature-scaled models.	`calibration_metrics.py`, `reliability_diagram_utils.py`
`seed_sensitivity_analysis.ipynb`	Evaluates model stability across random seeds using variance analysis and reliability diagrams.	`seed_sensitivity_utils.py`
(future) `distillation_inspection.ipynb`	Visualizes teacher–student transfer quality and calibration changes.	`03_distillation_experiment.py`
(future) `interpretability_analysis.ipynb`	Links SHAP or gradient-based attributions with calibration metrics.	`interpretability_utils.py`

⚙️ Environment Requirements

Python ≥ 3.10
PyTorch ≥ 2.0
Optuna or DEHB
MLflow 3.0
Pandas, NumPy, Matplotlib
(optional) Captum, SHAP for interpretability notebooks

Activate your environment:

```bash conda activate mugalab jupyter lab