Skip to content

Committor ts location - #3

Merged
carlocamilloni merged 3 commits into
mainfrom
committor-ts-location
Sep 18, 2026
Merged

carlocamilloni merged 3 commits into
mainfrom
committor-ts-location

Conversation

@carlocamilloni

Copy link
Copy Markdown
Member

No description provided.

The tilt fixes how much barrier is missing (r = p'/p) but not where it sits,
and the TS location depends on it.  For a narrow missing barrier on a model
isocommittor surface q* the corrected committor is
  q' = r q_m (U side),  1 - q' = r (1 - q_m) (F side),
exact for diffusive dynamics away from the barrier.

- maxcal_poisson: model committor from excursion outcomes (isotonic along Q),
  corrected committor, TS location (at the barrier when the jump straddles 1/2,
  otherwise on the U or F side); new options --barrier-q (default 0.5 = model
  TS, where the TS is lambda-independent), --tse-lams, --committor-bins,
  --tse-window; outputs committor.csv, ts_location.csv/.png,
  tse_committor_lam*_q*.csv
- tests: formula exact to 1e-3 on 1D committors of the test potential; the
  naive log-odds shift (barrier implicitly at Qf) misplaces the TS; sampled
  (slow) check that the predicted TS matches the truth within 0.03 in Q and a
  wrong barrier location does not; regtest checks the new outputs
- README: new section 2.10; restores list/code-block indentation that the
  previous spacing fix collapsed (math-block-only change this time)
- notebook: section 7 (TS location vs lambda and q*, against the truth)
- examples/walkthrough_ks_lambda.py: KS before/after tilt, lambda_min vs true lambda
…ween Q‡ and Qf

- maxcal_poisson: ks_heatmap() computes the median stitched KS p-value and CV
  on the (Q‡ scan grid) x lambda plane; heatmap.csv/.png with passing cells
  hatched, lambda_min(Q‡) overlaid, and Q‡ at or beyond the assumed barrier
  (from --barrier-q and the model committor) shaded as invalid; options
  --heatmap-nlam (0 disables) and --heatmap-lam-max; warning if the reference
  Q‡ is not before the assumed barrier; ts_location.csv flags
  ts_before_interface
- README: the corrected TS lies between Q‡ and Qf; decision-map section, with
  the finding that validity ends at the foot of the missing barrier (a rise of
  lambda_min is the observable signal) and that folding times cannot locate
  the barrier (same 3 kT barrier at 0.46/0.55/0.64: TS moves by 0.17, shapes
  indistinguishable)
- notebook: section 8, decision map with the true lambda(Q‡) of the ground
  truth overlaid and a slider for the assumed barrier location
- tests: heatmap shape/empty-pool unit test; regtest checks heatmap outputs
… version

A kernel started before a git pull/am keeps the old maxcal_poisson in memory,
which gave AttributeError: no attribute 'committor_frames' in section 7.  The
setup cell now reloads maxcal_poisson, maxcal_joint, langevin and synthetic,
prints which file is used, and raises a clear error if the module comes from
outside the repository or lacks current functions.
@carlocamilloni
carlocamilloni merged commit ae6f0e6 into main Sep 18, 2026
3 checks passed
@carlocamilloni
carlocamilloni deleted the committor-ts-location branch September 18, 2026 22:33
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants