update censoring notebook

[juanitorduz]

Use the recently added censored distribution + minor fix

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[github-actions]

Benchmark report

this PR update-censored-nb at a7de9450 vs baseline master at 07501df3

+ run time:     1 faster
- compile time: 1 slower, 0 faster

Significant changes (2)

                                ──────── run time ───────     ────── compile time ──────
  benchmark                     baseline   this PR      Δ     baseline   this PR       Δ
────────────────────────────────────────────────────────────────────────────────────────
+ lkj_cholesky_sample             4.3 ms    3.9 ms  -8.8%     830.1 ms  848.2 ms   +2.2%
- predictive_forward_sampling   551.5 ms  546.3 ms  -0.9%     114.9 ms  151.2 ms  +31.6%

_{Red is slower, green is faster; a row is coloured by the worse of its two columns. A delta in parentheses cleared the threshold on a measurement below the resolution floor, so it is shown without being called a change. † marks a benchmark that could not be compared — see below.}

Full results

distributions

                                 ───────── run time ────────     ────── compile time ─────
  benchmark                      baseline  this PR         Δ     baseline   this PR      Δ
──────────────────────────────────────────────────────────────────────────────────────────
  biject_to_constraints            3.1 ms   3.0 ms     -1.5%     197.6 ms  210.0 ms  +6.3%
  categorical_log_prob             1.4 ms   1.4 ms     +3.0%      57.4 ms   56.8 ms  -1.0%
  dirichlet_log_prob               406 µs   385 µs   (-5.1%)     420.1 ms  386.6 ms  -8.0%
  dirichlet_sample                42.8 ms  43.0 ms     +0.5%     592.4 ms  600.1 ms  +1.3%
  gamma_log_prob                   1.5 ms   1.4 ms     -0.8%       2.04 s    2.08 s  +2.2%
  gamma_sample                    19.4 ms  19.7 ms     +1.2%     526.2 ms  568.1 ms  +8.0%
+ lkj_cholesky_sample              4.3 ms   3.9 ms     -8.8%     830.1 ms  848.2 ms  +2.2%
  mixture_same_family_log_prob     568 µs   575 µs     +1.3%      68.4 ms   66.4 ms  -2.9%
  multivariate_normal_log_prob     403 µs   477 µs  (+18.6%)     480.5 ms  469.5 ms  -2.3%
  normal_log_prob                  449 µs   446 µs     -0.7%      49.5 ms   49.8 ms  +0.6%
  normal_sample                   20.2 ms  20.1 ms     -0.6%     151.5 ms  148.0 ms  -2.3%
  stick_breaking_transform         4.7 ms   4.5 ms     -2.4%     160.5 ms  162.4 ms  +1.1%
  student_t_log_prob               2.5 ms   2.5 ms     +0.0%      71.3 ms   72.1 ms  +1.1%
  truncated_normal_log_prob        529 µs   537 µs     +1.6%      44.2 ms   43.2 ms  -2.3%

handlers

                                  ──────── run time ───────     ────── compile time ──────
  benchmark                       baseline   this PR      Δ     baseline   this PR       Δ
──────────────────────────────────────────────────────────────────────────────────────────
  initialize_model_hierarchical    31.5 ms   31.3 ms  -0.6%       3.24 s    3.34 s   +3.1%
  log_density_hierarchical          2.4 ms    2.3 ms  -2.9%       1.26 s    1.19 s   -5.9%
  nested_handler_stack              976 µs    963 µs  -1.3%       781 µs    611 µs  -21.8%
  potential_energy_and_grad          18 µs     19 µs  +3.7%      78.4 ms   77.1 ms   -1.7%
- predictive_forward_sampling     551.5 ms  546.3 ms  -0.9%     114.9 ms  151.2 ms  +31.6%
  trace_seeded_model                530 µs    536 µs  +1.0%     450.9 ms  451.0 ms   +0.0%

mcmc

                             ──────── run time ───────     ───── compile time ─────
  benchmark                  baseline   this PR      Δ     baseline  this PR      Δ
───────────────────────────────────────────────────────────────────────────────────
  hmc_logistic_regression    589.8 ms  581.8 ms  -1.4%       2.82 s   2.79 s  -0.9%
  nuts_dense_mass_funnel     928.3 ms  924.1 ms  -0.5%       2.27 s   2.30 s  +1.4%
  nuts_eight_schools         923.7 ms  890.9 ms  -3.6%       2.33 s   2.31 s  -1.0%
  nuts_hierarchical_glm        3.25 s    3.24 s  -0.2%       3.97 s   4.23 s  +6.8%
  nuts_logistic_regression   869.8 ms  857.4 ms  -1.4%       2.79 s   3.02 s  +8.3%
  nuts_vectorized_chains       1.93 s    1.95 s  +1.3%       2.42 s   2.37 s  -1.8%

svi

                                             ──────── run time ───────     ───── compile time ─────
  benchmark                                  baseline   this PR      Δ     baseline  this PR      Δ
───────────────────────────────────────────────────────────────────────────────────────────────────
  svi_autodelta_map_logistic                 235.8 ms  243.6 ms  +3.3%       2.70 s   2.76 s  +2.5%
  svi_autodiagonalnormal_hierarchical        811.3 ms  813.6 ms  +0.3%       4.51 s   4.44 s  -1.5%
  svi_automultivariatenormal_eight_schools   592.5 ms  607.1 ms  +2.5%       3.90 s   4.04 s  +3.6%
  svi_autonormal_logistic                    596.3 ms  607.1 ms  +1.8%       3.05 s   3.10 s  +1.6%
  svi_multi_particle_elbo                      1.15 s    1.19 s  +3.6%       3.14 s   3.19 s  +1.8%
  svi_trace_mean_field_elbo                  986.0 ms    1.02 s  +3.0%       4.74 s   4.59 s  -3.2%

Methodology and environment

Each benchmark is set up untimed, then called once with the JAX caches cleared and several more times warm. Run is the fastest warm call; compile is the first call minus that, i.e. the tracing, lowering and XLA compilation the warm calls did not have to pay for.

Both refs were measured on the same runner over 2 interleaved round(s), taking the best observation per benchmark. A result is called neutral when it moves less than ±5% (run) or ±25% (compile), or when the measurement itself is under 1 ms (run) / 50 ms (compile) — a shared CI runner cannot resolve changes below that. Compile time gets the looser band because it is measured once per round rather than best-of-N, and swings by roughly 20% even between two runs of identical code. A delta shown in parentheses did clear its threshold, but on a measurement below the resolution floor, so it is reported without being called a change.

	baseline	this PR
ref	master	update-censored-nb
commit	07501df3	a7de9450
numpyro	0.21.0	0.21.0
jax	0.11.1	0.11.1
backend	cpu	cpu
python	3.14.7	3.14.7

Runner: Linux-6.17.0-1022-azure-x86_64-with-glibc2.39, 4 CPUs.

_{Produced by this benchmark run.}

[juanitorduz]

wdyt @Qazalbash ?

[Qazalbash]

wdyt @Qazalbash ?

I will soon look into and update you!

[Qazalbash]

The changes seem reasonable to me, except for some formatting-related issues in the notebook for latex.

Once @fehiepsi reviews it, we can merge this PR.