nirs4all.pipeline.dagml.structural_classification module

Serial native classification declarations, grouped scope preflight and capture.

nirs4all.pipeline.dagml.structural_classification.classification_output(graph: Mapping[str, Any], node_id: str, vocabulary: Mapping[str, Any]) → dict[str, Any]

Bind the label port; probability columns remain signed native auxiliaries.

nirs4all.pipeline.dagml.structural_classification.declared_classifier_topologies(steps: list[Any], splitter: Any) → list[list[Any]]

Validate public grammar without native state or recipe enumeration.

nirs4all.pipeline.dagml.structural_classification.is_classification_choice(steps: list[Any]) → bool

Recognize explicit native classifiers without consulting fitted state.

nirs4all.pipeline.dagml.structural_classification.lower_classifier_choices(steps: list[Any], splitter: Any, schemas: Mapping[str, Any], vocabulary: Mapping[str, Any]) → tuple[dict[str, Any], dict[str, Any], list[str]]

Lower each public declaration once; native compilation owns expansion.

nirs4all.pipeline.dagml.structural_classification.prepare_classification_structure(steps: list[Any], splitter: Any, dataset: Any, spec: DagMLTuningSpec, run_options: dict[str, Any], native: Any, *, require_space: bool = True) → dict[str, Any]

Sign Train-only label identities; native validates every actual train scope.

nirs4all.pipeline.dagml.structural_classification.run_fixed_classifier(pipeline: Any, dataset: Any, *, name: str, random_state: int | None, refit: bool) → Any

Preflight a fixed classifier before native CV/SELECT/REFIT, without HPO.

nirs4all.pipeline.dagml.structural_classification.train_selected_classification(prepared: dict[str, Any], evidence: dict[str, Any] | None, run_options: dict[str, Any]) → Any

Resolve the exact native sink and capture all its portable predecessors.

nirs4all.pipeline.dagml.structural_classification.validate_count_space(declaration: Any) → None

Require an integer proposal domain; native checks actual fold capacities.