[multi_hyper] Conversion to JAX and stylesheet check

[HumphreyYang]

This PR converts the simulation code to JAX and perform stylesheet check!

[github-actions]

📖 Netlify Preview Ready!

Preview URL: https://pr-657--sunny-cactus-210e3e.netlify.app (f4a251b)

📚 Changed Lecture Pages: multi_hyper

[jstac]

🤖 Status note for a future session — from a maintainer investigation on 2026-07-08 into why open-PR previews 404. Context only, not instructions.

Netlify preview: https://pr-657--sunny-cactus-210e3e.netlify.app/ currently returns 404.

Why previews are down (repo-wide findings)

1. This branch is stale — 166 commits behind main. A preview build compiles the whole site from this branch. This branch's lectures/house_auction.md still has unpinned !pip install prettytable, which now breaks on a wcwidth incompatibility. main fixed this on 2026-06-28 by pinning prettytable<3.18 (#939). This alone fails any rebuild of this branch until it's updated to main.

2. The arviz failure was a red herring — do NOT pin arviz or rewrite plotting. A 2026-07-07 rebuild also failed in ar1_bayes/ar1_turningpts with an arviz_plots figsize ValueError. That was a transient bug in an intermediate arviz-plots 1.x release, already fixed in arviz 1.2.0. Verified locally on a clean latest-stack venv: the real az.plot_trace(trace) cell (pymc + numpyro InferenceData) runs green. The lectures use only 1.x-compatible arviz APIs (plot_trace, summary, from_numpyro, compare).

Recommended first step for this PR

Update this branch to main (merge or rebase — pulls in #939 plus ~166 other commits), then let CI rebuild. On today's latest libraries the site builds clean, so the preview should return. house_auction is the known blocker; updating also picks up other since-merged fixes — rebuild and address any remaining per-lecture failures. Verify with:

curl -sI https://pr-657--sunny-cactus-210e3e.netlify.app/multi_hyper.html

---

This PR touches: multi_hyper.md. Last CI build: success@2025-10-26. Branch: 166 commits behind main as of 2026-07-08.

[jstac]

Thanks @HumphreyYang. I've reworked this branch rather than let it sit — here's the reasoning.

Dropping the JAX conversion. The sampling in this lecture is a single call to np.random.Generator.multivariate_hypergeometric, a library primitive that JAX has no equivalent for. Converting therefore meant hand-rolling the sampler as a lax.scan over individual draws, vmap'd across the sample — which is the rewriting a library call by hand anti-pattern in the style guide.

The hand-rolled version is correct — sample means match theory and every row sums to $n$ — but it's slower than NumPy at every size, measured on CPU, which the branch forced via jax_platform_name:

size	NumPy	this branch
10,000	0.004s	0.358s	87× slower
100,000	0.012s	0.481s	40× slower
1,000,000	0.153s	0.656s	4.3× slower
10,000,000 (what the lecture uses)	1.35s	3.69s	2.7× slower

There was also a subtler consequence. Reimplementing the PMF in log space with jax.scipy.special.gammaln runs in float32, so the first example returned

jax gammaln : 0.079575970768929

where the prose two lines above states the exact value 0.079575596816976. The reader would see the two disagree at the sixth significant figure.

Separately, !pip install jax at the top is something the style guide rules out — it can install jax[cpu] rather than the GPU configuration we build with, which comes from Docker and Actions.

What I kept. Your vectorization of the covariance matrix is a real improvement and doesn't need JAX at all, so it's now applied on top of main's NumPy implementation:

p = K_arr / N
Σ = n * (N - n) / (N - 1) * (np.diag(p) - np.outer(p, p))

That replaces the nested loop and its easily-missed trailing Σ[-1, -1] special case. I checked it against the old code — maximum absolute difference 2.2e-16.

Also kept: the section headings in sentence case, and {code-cell} python3 / {code-cell} ipython brought to ipython3 throughout.

I left the emphasis as it is on main. Most of the bolded items here — "balls", "color", "color blind", "binomial coefficients" — are terms being defined, and the style guide reserves bold for definitions with italic for emphasis, so the existing markup already matches.

I also merged current main (the branch was 220 commits behind, which was why the preview was 404ing) and marked the PR ready. It's now +31/−37 against main.

[github-actions]

📖 Netlify Preview Ready!

Preview URL: https://pr-657--sunny-cactus-210e3e.netlify.app

Commit: 67560aa

📚 Changed Lectures

• multi_hyper

---

Build Info

• Workflow: Build Project [using jupyter-book]

[jstac]

Thanks @HumphreyYang for working on this. Comments above are by Claude. Modifications are by Claude after reading the latest style guide. I'm doing whatever it says because I'm scared of making it angry. I'll merge when green.

[mmcky]

✅ Translation sync completed (zh-cn)

Target repo: QuantEcon/lecture-python.zh-cn
Translation PR: QuantEcon/lecture-python.zh-cn#241
Files synced (1):

• lectures/multi_hyper.md