`airsspy.volume_minsep_model`#

Lightweight baseline predictor for volume and minsep estimates.

Module Contents#

Classes#

`FeatureBuilder`	Build reusable formula and pair feature vectors.
`BaselineRegressor`	Ridge regressor with explicit feature standardization.
`BaselineFormulaPredictor`	Predict formula volume and minsep values from a baseline bundle.

Functions#

`split_name_for_formula`	Assign a deterministic split label to a reduced formula.
`fit_ridge_regressor`	Fit a ridge regressor in closed form.
`evaluate_regressor`	Return simple regression metrics.
`train_baseline_regressor`	Train a deterministic ridge model with validation alpha selection.
`save_baseline_bundle`	Write a JSON baseline bundle.
`load_baseline_bundle`	Load a JSON baseline bundle.
`train_baseline_models`	Train baseline volume and minsep regressors from a dataset JSON file.

Data#

PROPERTY_NAMES

API#

airsspy.volume_minsep_model.PROPERTY_NAMES#: (‘Z’, ‘X’, ‘atomic_mass’, ‘atomic_radius’, ‘atomic_radius_calculated’, ‘metallic_radius’, ‘average_i…

class airsspy.volume_minsep_model.FeatureBuilder[source]#

Build reusable formula and pair feature vectors.

element_list: tuple[str, ...]#: None

__post_init__() → None[source]#

classmethod from_formulas(formulas: collections.abc.Iterable[str]) → airsspy.volume_minsep_model.FeatureBuilder[source]#

property formula_feature_names: list[str]#

property pair_feature_names: list[str]#

formula_features(formula: str) → numpy.ndarray[source]#

pair_features(formula: str, pair_key: str) → numpy.ndarray[source]#

airsspy.volume_minsep_model.split_name_for_formula(formula: str, *, seed: int, train_ratio: float = 0.8, val_ratio: float = 0.1) → str[source]#: Assign a deterministic split label to a reduced formula.

class airsspy.volume_minsep_model.BaselineRegressor[source]#

Ridge regressor with explicit feature standardization.

alpha: float#: None

transform: str#: None

feature_mean: numpy.ndarray#: None

feature_scale: numpy.ndarray#: None

target_mean: float#: None

coefficients: numpy.ndarray#: None

intercept: float#: None

feature_names: list[str]#: None

predict_raw(features: numpy.ndarray) → numpy.ndarray[source]#

predict(features: numpy.ndarray) → numpy.ndarray[source]#

to_payload() → dict[source]#

classmethod from_payload(payload: dict) → airsspy.volume_minsep_model.BaselineRegressor[source]#

airsspy.volume_minsep_model.fit_ridge_regressor(features: numpy.ndarray, targets: numpy.ndarray, *, alpha: float, feature_names: list[str], sample_weight: numpy.ndarray | None = None, transform: str = 'log') → airsspy.volume_minsep_model.BaselineRegressor[source]#: Fit a ridge regressor in closed form.

airsspy.volume_minsep_model.evaluate_regressor(model: airsspy.volume_minsep_model.BaselineRegressor, features: numpy.ndarray, targets: numpy.ndarray) → dict[str, float][source]#: Return simple regression metrics.

airsspy.volume_minsep_model.train_baseline_regressor(items: collections.abc.Sequence[airsspy.volume_minsep_data.VolumeAggregate] | collections.abc.Sequence[airsspy.volume_minsep_data.PairAggregate], *, feature_fn: collections.abc.Callable[[object], numpy.ndarray], target_fn: collections.abc.Callable[[object], float], weight_fn: collections.abc.Callable[[object], float], feature_names: list[str], seed: int, alpha_grid: collections.abc.Sequence[float], max_rows: int | None = None) → tuple[airsspy.volume_minsep_model.BaselineRegressor, dict][source]#: Train a deterministic ridge model with validation alpha selection.

airsspy.volume_minsep_model.save_baseline_bundle(output_dir: str | pathlib.Path, *, builder: airsspy.volume_minsep_model.FeatureBuilder, volume_model: airsspy.volume_minsep_model.BaselineRegressor, minsep_model: airsspy.volume_minsep_model.BaselineRegressor, metrics: dict, dataset_summary: dict) → None[source]#: Write a JSON baseline bundle.

airsspy.volume_minsep_model.load_baseline_bundle(path: str | pathlib.Path) → tuple[airsspy.volume_minsep_model.FeatureBuilder, airsspy.volume_minsep_model.BaselineRegressor, airsspy.volume_minsep_model.BaselineRegressor, dict][source]#: Load a JSON baseline bundle.

airsspy.volume_minsep_model.train_baseline_models(*, dataset_path: str | pathlib.Path, output_dir: str | pathlib.Path, seed: int = 17, pair_min_count: int = 1, volume_alpha_grid: collections.abc.Sequence[float] = (1e-06, 0.0001, 0.01, 1.0, 100.0), minsep_alpha_grid: collections.abc.Sequence[float] = (1e-06, 0.0001, 0.01, 1.0, 100.0), max_volume_rows: int | None = None, max_pair_rows: int | None = None) → dict[source]#: Train baseline volume and minsep regressors from a dataset JSON file.

class airsspy.volume_minsep_model.BaselineFormulaPredictor[source]#

Predict formula volume and minsep values from a baseline bundle.

bundle_path: str | pathlib.Path#: None

__post_init__() → None[source]#

predict(formula: str) → dict[source]#

predict_many(formulas: collections.abc.Sequence[str]) → list[dict][source]#: Predict many formulas in order.

airsspy.volume_minsep_model#