1 reporting.score_floor

reporting.score_floor

Source-bound readout for the matched score-scale-floor experiment.

1.1 Classes

Name Description
ScoreFloorReadout Normalized lifecycle, observation, and source-root comparison tables.

1.1.1 ScoreFloorReadout

reporting.score_floor.ScoreFloorReadout(
    source,
    stores,
    steps,
    seed_profiles,
    seed_prefix_effects,
    root_profiles,
    root_prefix_effects,
)

Normalized lifecycle, observation, and source-root comparison tables.

steps contains repeated observations from factual trajectories. It is useful for mechanics and missingness, but those rows are correlated within stores and roots. seed_prefix_effects retains one row per proposal seed and requested prefix, including unsupported endpoints as missing values. root_profiles and root_prefix_effects then reduce matched proposal seeds to one row per independent source root before comparing profiles.

1.2 Functions

Name Description
export_score_floor_readout Export source-bound JSON, CSV tables, and root-unit Seaborn SVGs.
read_score_floor_experiment Read the frozen matched experiment without modifying its stores.
score_floor_figures Build the canonical root-unit figures shared by exports and applications.
score_floor_reducer_identity Bind the readout to its reducer and semantic inspection dependencies.
score_floor_snapshot_identity Return the lightweight retained-result key for the default readout.
score_floor_summary Return completed-store root reductions and explicit limitations.

1.2.1 export_score_floor_readout

reporting.score_floor.export_score_floor_readout(
    plan_path,
    output_dir,
    *,
    azimuth_bins=_DEFAULT_AZIMUTH_BINS,
    elevation_bins=_DEFAULT_ELEVATION_BINS,
    min_intervening_steps=_DEFAULT_MIN_INTERVENING_STEPS,
)

Export source-bound JSON, CSV tables, and root-unit Seaborn SVGs.

1.2.2 read_score_floor_experiment

reporting.score_floor.read_score_floor_experiment(
    plan_path,
    *,
    azimuth_bins=_DEFAULT_AZIMUTH_BINS,
    elevation_bins=_DEFAULT_ELEVATION_BINS,
    min_intervening_steps=_DEFAULT_MIN_INTERVENING_STEPS,
)

Read the frozen matched experiment without modifying its stores.

read_recipe_experiment_report admits immutable successful stores, which are then reopened through the canonical read-only rollout reader. Recorded-selection validation defines actor-valid support and stochastic policy mechanics, while selected-motion frames define physical and target-relative path summaries.

Per-step rows are observations, not independent experimental units. For a root :math:r, profile :math:a, and matched proposal seed :math:s, the endpoint at prefix :math:h is the stored cumulative root gain :math:G_{r,a,s,h}. The root-paired effect is

.. math::

{r,h}={sS_{r,h}} (G_{r,,s,h}-G_{r,,s,h}),

where S contains only completed stores with both observed endpoints. Queue-wide attempt accounting is deliberately absent: missing, failed, and unreached attempts are neither counted nor imputed as zero.

1.2.3 score_floor_figures

reporting.score_floor.score_floor_figures(readout)

Build the canonical root-unit figures shared by exports and applications.

1.2.4 score_floor_reducer_identity

reporting.score_floor.score_floor_reducer_identity()

Bind the readout to its reducer and semantic inspection dependencies.

1.2.5 score_floor_snapshot_identity

reporting.score_floor.score_floor_snapshot_identity(plan_path)

Return the lightweight retained-result key for the default readout.

1.2.6 score_floor_summary

reporting.score_floor.score_floor_summary(readout)

Return completed-store root reductions and explicit limitations.

Back to top