Discrete Feature Helper

GitHub Link to Code.

Helper for discrete feature data preparation.

Provides data conversion for discrete features to prepare them for visualization by mapping categorical values to integer positions.

class mdxplain.plots.helper.discrete_feature_helper.DiscreteFeatureHelper

Helper for discrete feature data preparation.

Provides data conversion for discrete features to prepare them for visualization by mapping categorical values to integer positions.

Examples

>>> # Convert character data to positions
>>> data = np.array(['H', 'E', 'C', 'H'])
>>> mapping = {'H': 0, 'E': 1, 'C': 2}
>>> positions = DiscreteFeatureHelper.prepare_discrete_data(data, mapping)
>>> print(positions)
[0 1 2 0]

static prepare_discrete_data(data: ndarray, value_to_position: Dict) → ndarray

Convert discrete data to integer positions for plotting.

Maps categorical data values (integers or characters) to sequential integer positions suitable for visualization.

Parameters

datanp.ndarray: Original data (integers or character strings)
value_to_positiondict: Mapping from data values to plot positions

Returns

np.ndarray: Data converted to integer positions

Examples

>>> # Integer data (no conversion needed)
>>> data = np.array([0, 1, 2, 0, 1])
>>> mapping = {0: 0, 1: 1, 2: 2}
>>> positions = DiscreteFeatureHelper.prepare_discrete_data(data, mapping)
>>> print(positions)
[0 1 2 0 1]

>>> # Character data (needs conversion)
>>> data = np.array(['H', 'E', 'C', 'H', 'E'])
>>> mapping = {'H': 0, 'E': 1, 'C': 2}
>>> positions = DiscreteFeatureHelper.prepare_discrete_data(data, mapping)
>>> print(positions)
[0 1 2 0 1]

Notes

Values not present in value_to_position are returned as -1.

static build_axis_config(selector_data: Dict[str, ndarray] | None = None, viz: Dict[str, Any] | None = None, long_labels: bool = False, x_padding: float = 0.3, fallback_from_data: bool = False) → Dict[str, Any]

Build axis configuration for discrete plotting.

Parameters

selector_dataDict[str, np.ndarray], optional: Selector values used for optional data-driven fallback.
vizDict[str, Any], optional: Visualization metadata (expects optional tick_labels).
long_labelsbool, default=False: If True, uses long tick labels from metadata when available.
x_paddingfloat, default=0.3: Horizontal padding applied to x-limits.
fallback_from_databool, default=False: If True and metadata has no tick labels, build positions from unique observed values in selector_data. If False, binary fallback.

Returns

Dict[str, Any]

Axis configuration with keys:

positions: np.ndarray
value_to_position: Dict[Any, int]
tick_labels: List[str]
xlim: tuple(float, float)

static calculate_discrete_probabilities(data: ndarray, value_to_position: Dict[Any, int], n_positions: int) → ndarray

Convert discrete samples into per-position probabilities.

Parameters

datanp.ndarray: Raw discrete samples for one selector.
value_to_positionDict[Any, int]: Mapping from raw discrete values to axis position indices.
n_positionsint: Total number of discrete axis positions.

Returns

np.ndarray: Probability vector of length n_positions.

Notes

Samples that cannot be mapped to a valid position are ignored.