diff options
Diffstat (limited to 'notes/36_evidence_ledger.md')
| -rw-r--r-- | notes/36_evidence_ledger.md | 7 |
1 files changed, 5 insertions, 2 deletions
diff --git a/notes/36_evidence_ledger.md b/notes/36_evidence_ledger.md index 1028f48..9807e06 100644 --- a/notes/36_evidence_ledger.md +++ b/notes/36_evidence_ledger.md @@ -66,8 +66,11 @@ standard lazy approximation whose breakdown is *measured* (item 3c below). ### (b) Exact at t=0 (T5 + T6) - Mean theorem, 2-hidden-layer, real FA backward, 6 widths x 6 inits x 512 - draws: `E_B[e_0] = 0.4378 +/- 0.008` vs theory `0.4403`; per-width errors - 0.0003-0.008 (note 29, `actual_fa_initial_operator_moments`). + draws: no-fit prediction within Monte-Carlo error across all 36 rows, + **max |error| <= 0.008**; predicted erosion spans 0.15-0.45 (wide-range + calibration, not one point). Full table in note 42 §6 + (`actual_fa_initial_operator_moments`). The old "0.4378/0.4403" headline is + NOT in the current CSV -- retired, see note 37 §4. - Independent t=0 anchor from the (superseded) recovery probe: 0.4026+/-0.0066 vs 0.4071 (run log, note 37). - Full distribution (T6), **sampler = real FA backward after note-37 fix**: |