Gaussian Process Summer School lectures

Lecture site for the Gaussian Process Summer School, published at mlatcl.github.io/gpss.

Talk sources live in _lamd/ and are compiled with LaMD (maketalk). Shared narrative and code includes come from lawrennd/snippets (see snippetsdir in _lamd/_lamd.yml). Generated lecture HTML, notebooks, and Reveal slides are committed under _lectures/, _notebooks/, and slides/ for GitHub Pages (gh-pages).

Local site preview

bundle install
bundle exec jekyll serve
# Visit http://localhost:4000/gpss/

Rebuild a lecture

Requires a local LaMD install and paths in _lamd/_lamd.yml (snippetsdir, bibdir, macrosdir). Prefer the LaMD virtualenv on PATH so mdfield / maketalk resolve correctly:

export PATH="$HOME/lawrennd/lamd/.venv/bin:$PATH"
cd _lamd
maketalk deep-gps-I.md
maketalk deep-gps-II.md
# or: ./compile.sh   # rebuilds every stub in lectures.csv

Notebook runtime dependencies

Deep GP notebooks install teaching helpers via LaMD \installcode{} cells, typically:

• GPy
• PyDeepGP (import deepgp)
• mlai (includes mlai.deepgp_tutorial: initialize, staged optimize, visualize, visualize_pinball)

Pinball and related Deep GP teaching methods live in lawrennd/mlai, not in PyDeepGP itself; lecture setup monkey-patches them onto deepgp.DeepGP.

Layout

Path	Role
_lamd/*.md	Lecture sources
_lamd/_lamd.yml	LaMD paths and talk metadata
_people/people.yml	Speaker metadata for talks
_lectures/	Generated lecture HTML (Jekyll)
_notebooks/	Generated Jupyter notebooks
slides/	Reveal.js slides and diagrams/