Skip to content

[archive] Piecewise/LR landmark observation model (unused)#635

Closed
efahnestock wants to merge 1 commit into
mainfrom
archive/piecewise-landmark-obs-model
Closed

[archive] Piecewise/LR landmark observation model (unused)#635
efahnestock wants to merge 1 commit into
mainfrom
archive/piecewise-landmark-obs-model

Conversation

@efahnestock

@efahnestock efahnestock commented Jul 2, 2026

Copy link
Copy Markdown
Collaborator

Archival PR — intended to be closed unmerged, not reviewed/merged.

Preserves the piecewise likelihood-ratio landmark observation stream explored as an alternative to the half-normal landmark model. Performed better in some settings and worse in others

  • SafaPlusPiecewiseLandmarkAggregator (+ config/dispatch/tests) in filter/adaptive_aggregators.py
  • scripts/calibrate_landmark_piecewise.py, scripts/plot_piecewise_calibration.py
  • ZOI precursor scripts/calibrate_landmark_zoi.py and scripts/residual_likelihood_analysis.py

The piecewise likelihood-ratio landmark observation stream that was explored as
an alternative to the half-normal landmark model, but not adopted:

- SafaPlusPiecewiseLandmarkAggregator (+ config, dispatch, tests) in
  filter/adaptive_aggregators.py
- scripts: calibrate_landmark_piecewise.py, plot_piecewise_calibration.py,
  the ZOI precursor calibrate_landmark_zoi.py, and residual_likelihood_analysis.py

Not used: on Norway the normalized-residual log-LR rewards spurious low-r
matches while the true patch sits in the miss bin, diverging the filter at full
strength. Archived for reference; this PR is intended to be closed unmerged.
@efahnestock efahnestock closed this Jul 2, 2026
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.

1 participant