phenotypic.tune.ScreeningResult#

class phenotypic.tune.ScreeningResult(winner: ~phenotypic.tune._study_store.Trial | None, frozen: dict[str, ~typing.Any] = <factory>, method: ~typing.Literal['fanova', 'rf-permutation'] = 'rf-permutation', interactions_estimated: bool = False, screened: bool = False, freeze_flagged: bool = False, reduced_space: ~phenotypic.tune._search_space._space.SearchSpace | None = None, recommendation: str = '')[source]#

Bases: object

The outcome of a (possibly two-round) screening run.

Parameters:

  • winner (Trial | None) – The best held-out Trial across both rounds (None only if nothing succeeded).

  • frozen (dict[str, Any]) – {frozen_key: pinned_value} — empty when no freeze occurred.

  • method (Literal['fanova', 'rf-permutation']) – The importance method used (fANOVA vs RF-permutation).

  • interactions_estimated (bool) – Whether method accounts for interactions.

  • screened (bool) – Whether the focused (freeze) round actually ran.

  • freeze_flagged (bool) – True when the focused round underperformed explore on held-out and the freeze was flagged as likely-bad (G3).

  • reduced_space (SearchSpace | None) – The reduced (frozen) space the focused round searched, or None when no freeze occurred.

  • recommendation (str) – A plain-English next step for the operator.

Methods

__init__

Attributes

freeze_flagged

interactions_estimated

method

recommendation

reduced_space

screened

winner

frozen

winner: Trial | None#
frozen: dict[str, Any]#
method: Literal['fanova', 'rf-permutation'] = 'rf-permutation'#
interactions_estimated: bool = False#
screened: bool = False#
freeze_flagged: bool = False#
reduced_space: SearchSpace | None = None#
recommendation: str = ''#