phenotypic.tune.compute_param_importance

phenotypic.tune.compute_param_importance(store: StudyStore, *, random_state: int = 0, objective: str | None = None) → dict[str, float]

Rank tuned parameters by importance against the objective (the dict view).

The back-compat, dict-returning façade over compute_param_importance_report(): it dispatches the same way (native fANOVA when the store offers it, RandomForest + permutation otherwise) but discards the method / honesty metadata.

Parameters:

• store (StudyStore) – The study store of completed trials.

• random_state (int) – Seed for the forest + permutation (reproducibility); only used on the RandomForest fallback path.

• objective (str | None) – The named multi-objective to rank against (plan §0a sidecar). None (default) ranks against Trial.score — the unchanged single-objective path. A name ranks against Trial.objectives[name], skipping trials that lack the objective.

Returns:

{param_key: importance} sorted descending. Empty when fewer than two usable trials (nothing to fit).

Return type:

dict[str, float]