Cite published JMLR version of the WALNUTS paper - concept-collection/walnuts-interactive

concept-collection / walnuts-interactive

Cite published JMLR version of the WALNUTS paper

Jeremy Magland <jmagland@flatironinstitute.org> committed 1 month ago commit 38237c781f39 parent dfe6697 Browse files

3 changed files+6−3

CLAUDE.mdmodified+2−1View file

@@ -41,7 +41,8 @@ orbit's positions, and packs `px/py/seg` + the start and selected draw.

4141

4242 `helpers/walnuts.m` and its building blocks (`extend_orbit_forward/backward`,

4343 `micro`, `leapfrog`, `u_turn`/`sub_u_turn`, `p_micro`/`pmf_p_micro`) are **Nawaf

44-Bou-Rabee's reference MATLAB implementation**, used as provided (paper

44+Bou-Rabee's reference MATLAB implementation**, used as provided (paper: JMLR

45+27(113):1–64, 2026, https://jmlr.org/papers/v27/25-1452.html; preprint

4546 arXiv:2506.18746). It's the orbit-based formulation: sample momentum, grow the

4647 orbit by doubling forward/backward, refine the leapfrog step within each step so

4748 the energy variation ≤ `δ`, terminate on a U-turn, and pick a draw from the

README.mdmodified+2−1View file

@@ -48,7 +48,8 @@ orbits; the script replies with `sendEventToHTMLSource`.

4848

4949 - **Algorithm & reference MATLAB implementation:** N. Bou-Rabee, B. Carpenter,

5050 T. S. Kleppe, and S. Liu, *The within-orbit adaptive leapfrog no-U-turn

51- sampler*, [arXiv:2506.18746](https://arxiv.org/abs/2506.18746) (2025). The

51+ sampler*, [JMLR 27(113):1–64](https://jmlr.org/papers/v27/25-1452.html)

52+ (2026); preprint [arXiv:2506.18746](https://arxiv.org/abs/2506.18746). The

5253 `helpers/` sampler (`walnuts.m` + its building blocks) is Nawaf Bou-Rabee's

5354 reference MATLAB code, used here as provided. Reference C++ at

5455 [flatironinstitute/walnuts](https://github.com/flatironinstitute/walnuts).

helpers/walnuts.mmodified+2−1View file

@@ -1,7 +1,8 @@

11 function [theta_tilde, T, O] = walnuts(logmu, gradlogmu, theta, h, i_max, delta)

22 %WALNUTS One WALNUTS transition (within-orbit adaptive leapfrog NUTS).

33 % Reference MATLAB implementation aligned with the paper

4-% arXiv:2506.18746 (Bou-Rabee, Carpenter, Kleppe, Liu). LOGMU / GRADLOGMU are

4+% (Bou-Rabee, Carpenter, Kleppe, Liu, JMLR 27(113):1-64, 2026,

5+% https://jmlr.org/papers/v27/25-1452.html). LOGMU / GRADLOGMU are

56 % function handles for the target's log density and gradient. Returns the

67 % selected draw THETA_TILDE, the orbit length T, and the orbit O (a cell of

78 % {theta, rho} states) — O is exposed only for the figure's movie and does not