f62098bReset cache: discard draws made before auto-thinningJeremy Magland 1# Cached entropy estimates
3285002estimates: run 8 from efed18agithub-actions[bot] 3Machine-generated by [`.github/workflows/estimates.yml`][wf] on `main` — do not
4edit by hand. Every dispatch runs the whole grid again with fresh, independent
5pasts, appends one record per cell to `runs.jsonl`, and rebuilds
6`estimates.json` from the full log, so the means tighten run after run.
8- **`runs.jsonl`** — append-only, one JSON object per (cell, dispatch): the
9 cell's parameters, the per-past values with their sum and sum of squares,
10 the per-past resolved Gibbs thinning and probe autocorrelation times
11 (`thin='auto'` matches thinning to each chain's measured mixing), the
12 seed, and the code commit that produced them.
13- **`estimates.json`** — pooled `mean` and `se` per cell over every past ever
14 drawn for it, alongside the analytic `predicted` rate for comparison.
16The estimand is H(z_(M+1) | z_1..z_M) in bits/sample — an upper bound on the
17entropy rate of z that tightens as the conditioning window M grows.
19Fetch it from a browser or a script (raw.githubusercontent.com sends
20`access-control-allow-origin: *`):
22 https://raw.githubusercontent.com/concept-collection/timeseries-entropy/estimates/estimates.json
2e15c7destimates: run 72 from efed18agithub-actions[bot] 24## Current state (64 runs, generated 2026-08-08T19:36:28Z)
26| cell | sigma | M | mean | se | pasts | predicted |
27| --- | --- | --- | --- | --- | --- | --- |
2e15c7destimates: run 72 from efed18agithub-actions[bot] 28| bp300-6000_s1 | 1 | 512 | 0.9340 | 0.0053 | 512 | 0.9513 |
29| bp300-6000_s16 | 16 | 512 | 2.5454 | 0.0100 | 512 | 2.5680 |
30| bp300-6000_s2 | 2 | 512 | 1.3113 | 0.0052 | 512 | 1.3319 |
31| bp300-6000_s32 | 32 | 512 | 2.9589 | 0.0192 | 512 | 2.9938 |
32| bp300-6000_s4 | 4 | 512 | 1.7209 | 0.0059 | 512 | 1.7342 |
33| bp300-6000_s8 | 8 | 512 | 2.1352 | 0.0100 | 512 | 2.1476 |
34| lp3000_s1 | 1 | 512 | 0.4764 | 0.0104 | 512 | 0.6166 |
35| lp3000_s16 | 16 | 512 | 1.4512 | 0.0083 | 512 | 1.4888 |
36| lp3000_s2 | 2 | 512 | 0.7558 | 0.0122 | 512 | 0.8251 |
37| lp3000_s32 | 32 | 512 | 1.6588 | 0.0089 | 512 | 1.7179 |
38| lp3000_s4 | 4 | 512 | 0.9884 | 0.0084 | 512 | 1.0415 |
39| lp3000_s8 | 8 | 512 | 1.2248 | 0.0096 | 512 | 1.2632 |
40| ma8_s1 | 1 | 512 | 0.4768 | 0.0092 | 512 | 0.6100 |
41| ma8_s16 | 16 | 512 | 3.1876 | 0.0121 | 512 | 3.1777 |
42| ma8_s2 | 2 | 512 | 0.9735 | 0.0045 | 512 | 0.9825 |
43| ma8_s32 | 32 | 512 | 4.0825 | 0.0104 | 512 | 4.1125 |
44| ma8_s4 | 4 | 512 | 1.5590 | 0.0048 | 512 | 1.5529 |
45| ma8_s8 | 8 | 512 | 2.3027 | 0.0067 | 512 | 2.3064 |
46| none_s1 | 1 | 512 | 2.1068 | 0.0034 | 512 | 2.1048 |
47| none_s16 | 16 | 512 | 6.0359 | 0.0226 | 512 | 6.0473 |
48| none_s2 | 2 | 512 | 3.0685 | 0.0043 | 512 | 3.0620 |
49| none_s32 | 32 | 512 | 7.0872 | 0.0478 | 512 | 7.0472 |
50| none_s4 | 4 | 512 | 4.0385 | 0.0061 | 512 | 4.0508 |
51| none_s8 | 8 | 512 | 5.0678 | 0.0141 | 512 | 5.0480 |
53[wf]: https://github.com/concept-collection/timeseries-entropy/blob/main/.github/workflows/estimates.yml