Teaching materials for an intensive day on hidden Markov models and mixed-state presentations, transformer belief geometry, and Hankel/PSR/WFA representations.
For co-presenters:
- Read
DAY_PLAN.mdfor the schedule and pacing choices. - Read
NOTEBOOK_OUTLINES.mdfor the route through each exercise block. - Read
HANDOFF.mdfor presenter notes, spoiler gates, and the current verification state. - Present the offline HTML decks from
slides/dist/. PressSfor speaker notes andOfor the slide overview. Matching PDFs are included as handouts and projection fallbacks.
For participants, distribute the three *_exercises.ipynb files in
notebooks/. Keep the matching *_solutions.ipynb files with the
teaching team until each reveal.
| Topic | Student notebook | Worked solutions |
|---|---|---|
| HMMs and mixed-state presentations | 01_hmms_msps_exercises.ipynb |
01_hmms_msps_solutions.ipynb |
| Transformer belief geometry | 02_transformer_belief_geometry_exercises.ipynb |
02_transformer_belief_geometry_solutions.ipynb |
| Hankel matrices, PSRs, and WFA reconstruction | 03_hankel_psr_wfa_exercises.ipynb |
03_hankel_psr_wfa_solutions.ipynb |
The notebooks contain embedded outputs and can be read without rerunning them. Running their supplied cells requires Python, NumPy, and Matplotlib.
notebooks/— generated exercise and solution notebooks, source builder, and validatorsslides/— Quarto sources, build tooling, offline presenter HTML, and PDF handoutsXavOriginal/— private source/reference material; do not distribute this folder to participantsPROJECT_CONTEXT.md— project goals and provenance
From the repository root:
python notebooks/build_notebooks.py --check
python notebooks/validate_notebooks.py
python slides/validate_slides.pyTo rebuild every notebook and slide artifact, including the offline decks and PDFs:
python slides/build_slides.pySee notebooks/README.md and
slides/README.md for the full editing and presentation
workflow.