eegdash.features.kinds

Feature Channel-processing Kinds.

This module defines the fundamental feature-processing kinds and the logic to map raw arrays to named features.

The module provides the classes:

• UnivariateFeature

• BivariateFeature

• MultivariateFeature

Classes

BivariateFeature(*args[, channel_pair_format])	Feature kind for operations on pairs of channels.
MultivariateFeature()	Logic wrapper for features that operate on one or more EEG channels.
UnivariateFeature()	Feature kind for operations applied to each channel independently.

class eegdash.features.kinds.BivariateFeature(*args, channel_pair_format: str | None = None)

Bases: MultivariateFeature

Feature kind for operations on pairs of channels.

Designed for undirected relationship measures between two signals.

Parameters:

channel_pair_format (str) – A format string used to create feature names from pairs of channel names. Default is “{}<>{}” for undirected bivariate features or “{}->{}” for directed bivariate features.

feature_channel_names(_metadata: dict) → list[str]

Generate feature names for each unique pair of channels.

Parameters:

_metadata (dict) – A dictionary of record and batch metadata.

Returns:

Formatted strings representing channel pairs (e.g., ‘F3<>F4’).

Return type:

list of str

class eegdash.features.kinds.MultivariateFeature

Bases: object

Logic wrapper for features that operate on one or more EEG channels.

This class defines the logic for mapping raw numerical results into structured, named dictionaries. It determines the “kind” of a feature (e.g., univariate, bivariate) and handles the association of feature values with specific channels or channel groupings.

Notes

Subclasses should override feature_channel_names() to define specific naming conventions for the extracted features.

feature_channel_names(_metadata: dict) → list[str]

Generate feature-specific names based on input channels.

Parameters:

_metadata (dict) – A dictionary of record and batch metadata.

Returns:

A list of strings defining the naming for each output feature. Returns an empty list in the base implementation.

Return type:

list of str

class eegdash.features.kinds.UnivariateFeature

Bases: MultivariateFeature

Feature kind for operations applied to each channel independently.

Used when a single feature value is produced per channel.

feature_channel_names(_metadata: dict) → list[str]

Return the channel names themselves as feature names.

Parameters:

_metadata (dict) – A dictionary of record and batch metadata.

Returns:

A list of channel names.

Return type:

list of str