[harrison_kreps] Update lecture

[xuanguang-li]

This pull request updates the Harrison-Kreps model lecture to use JAX and improves code consistency and formatting throughout the document.

• Replaced numpy (np) and scipy.linalg (la) imports and usages with jax.numpy (jnp) equivalents throughout code cells, but kept for and while loops for simplicity.

• Formatting and terminology improvements: standardized section and subsection titles, and improved consistency in emphasis (e.g., italicizing instead of bolding, fixing typos in comments and variable names).

[github-actions]

📖 Netlify Preview Ready!

Preview URL: https://pr-763--sunny-cactus-210e3e.netlify.app (f4ecd4b)

📚 Changed Lecture Pages: harrison_kreps

[github-actions]

📖 Netlify Preview Ready!

Preview URL: https://pr-763--sunny-cactus-210e3e.netlify.app

Commit: 5dc8bd4

📚 Changed Lectures

• harrison_kreps

---

Build Info

• Workflow: Build Project [using jupyter-book]

[jstac]

Thanks @xuanguang-li , much appreciated.

I've reviewed the PR.

Actually I don't think switching to JAX is justified in this lecture.

In the computations of the stationary distribution at the start, we will be using NumPy internally, since that's all qe.MarkovChain can work with.

In the later computation, with the for loop, JAX is probably slower, since the array computations are very low dimensional and the main task is iterative.

I think it might be best to close this PR.

@HumphreyYang or @mmcky, what do you think? If we do close this, let's tick it off or remove it from the meta issue, so we know not to try again.

[HumphreyYang]

Many thanks @jstac, I agree!

@xuanguang-li also has some other valuable edits in this PR so I can revert the JAX edits and keep others.

[xuanguang-li]

Thanks, @jstac and @HumphreyYang, for your comment.

I agree with your points. JAX usage is minimal in this lecture. Since the equilibrium computation isn't computationally heavy, I kept the for and while loops unchanged, so JAX is mainly for doing matrix algebra.

[jstac]

Thanks @xuanguang-li and @HumphreyYang

@HumphreyYang , it would be great if you could revert the jax edits.

There are lots of edge cases and in this lecture I think we are making the right decision.

[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-763--sunny-cactus-210e3e.netlify.app/ is currently live — but it's a lucky survivor. It was built 2026-06-05, before the breakages below, and simply hasn't been rebuilt since. Any rebuild will fail until the branch is updated (see below).

Why previews are down (repo-wide findings)

1. This branch is stale — 39 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 ~39 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-763--sunny-cactus-210e3e.netlify.app/harrison_kreps.html

---

This PR touches: harrison_kreps.md. Last CI build: success@2026-06-05. Branch: 39 commits behind main as of 2026-07-08.

[jstac]

Following up on the decision from May — @jstac concluded that JAX isn't justified here (the stationary distributions go through qe.MarkovChain, which is NumPy internally, and the remaining work is a low-dimensional iterative loop where JAX would likely be slower), @HumphreyYang agreed and offered to revert the JAX edits while keeping the rest, and @xuanguang-li agreed too. That last step never got finished, so I've done it now.

Where things stood. Humphrey's commits on 2026-06-05 had already removed the JAX changes — the branch head contained no jnp at all. What remained was a mix: the prose improvements, plus a set of reversions the branch was still carrying against main — the jupytext format_version/jupytext_version fields dropped, {code-cell} ipython3 back to {code-cell} ipython, :tags: [hide-output] back to the older ----delimited form, and the trailing newline removed. Those came from the branch being 51 commits behind main, not from anything anyone did deliberately.

What I did. Merged current main, resolved harrison_kreps.md in its favour to clear those reversions, then reapplied the prose work on top:

• eleven section and subsection headings to sentence case, per the style guide
• typos: "themsevles" → "themselves", "agaub" → "agent"
• bold → italic in the three places where the text is emphasis rather than a definition, which is the convention in the style guide
• the missing ## Exercises heading before the exercise block

The PR is now +19/−17 in one file.

One thing worth a second look, @jstac. The branch changed the $\check p(s)$ minimum equation from

$$P_a(s,0),\check p(0) + P_a(s,1),(1 + \check p(1))$$

to P_a(s,1) on the first term. I think that's the wrong direction: the first term is the probability of moving to state 0 times $\check p(0)$, and equation hakr2 above has it as P_a(s,0). What the branch was reacting to is real, though — the marginal-investor condition a few lines below hakr2 does read

$$P_a(s,1),\bar p(0) + P_a(s,1),(1 + \bar p(1)) < P_b(s,1),\bar p(0) + P_b(s,1),(1 + \bar p(1))$$

with P_a(s,1) and P_b(s,1) on the first term, which is inconsistent with hakr2 directly above it. So I've left the minimum equation as it was and fixed that condition to P_a(s,0) / P_b(s,0) instead. This is the one place I've changed mathematics rather than prose, so it's worth confirming before merge.

Thanks @xuanguang-li — the heading, typo and emphasis fixes are all worth having, and the equation inconsistency was a good catch even though the fix went the other way.

[mmcky]

✅ Translation sync completed (zh-cn)

Target repo: QuantEcon/lecture-python.zh-cn
Translation PR: QuantEcon/lecture-python.zh-cn#239
Files synced (1):

• lectures/harrison_kreps.md