eegdash.features.feature_bank.pick#

Channel-picking feature preprocessors#

This module provides the ability to pick specific channels or channel pairs for further processing.

Data Shape Convention#

By default, this module follows a Channel-panultimate convention:

  • Input: (..., channel, :)

  • Output: same as input

The choice of the channel dimension can be adjusted using the axis parameter.

Functions

pick_channel_pairs_preprocessor(*x, pairs, ...)

Pick a subset of channel pairs for further processing steps.

pick_channels_preprocessor(*x, channels, ...)

Pick a subset of channels for further processing steps.
eegdash.features.feature_bank.pick.pick_channel_pairs_preprocessor(*x, pairs: Iterable[Tuple[str, str]], _metadata: dict, index: int | Iterable[int] | None = -1, c_index: int | Iterable[int] | None = None, x_index: int | Iterable[int] | None = None, y_index: int | Iterable[int] | None = None, axis: int = -2, c_axis: int = -2)[source]

Pick a subset of channel pairs for further processing steps.

Must follow a preprocessor decorated with channel_pairer (or channel_directed_pairer).

Parameters:

  • *x (tuple[ndarray]) – Input batch.

  • pairs (Iterable[str]) – A list of channel pairs to pick.

  • index (int | Iterable[int]) – The index (or indices) of the input ndarray[s] to pick channel pairs from. Default is -1.

  • c_index (int | Iterable[int]) – The index (or indices) of the input ndarray[s] to pick channels from. Default is [].

  • x_index (int | Iterable[int]) – The index (or indices) of the input ndarray[s] to pick pair-first channels from. Default is [].

  • y_index (int | Iterable[int]) – The index (or indices) of the input ndarray[s] to pick pair-second channels from. Default is [].

  • axis (int) – The channel pairs axis of the input batch at index index. Default is -2.

  • c_axis (int) – The channels axis of the input batch at index c_index or x_index or y_index. Default is -2.

Returns:

  • *ndarray – Sliced input batch containing only the picked channels.

  • _metadata (dict) – Updated metadata dictionary.

Note

Picking by index pair, e.g., x[i, j], is not directly supported because the result may not be an numpy.ndarray. It is preferred to use a pair axis. It is possible, however, to pick just by x_index with c_axis=0, then pick again just by y_index with c_index=1 (or vice versa) to effectively pick the indices intersection of such an numpy.ndarray.

eegdash.features.feature_bank.pick.pick_channels_preprocessor(*x, channels: Iterable[str], _metadata: dict, index: int | Iterable[int] = -1, axis: int = -2)[source]

Pick a subset of channels for further processing steps.

Parameters:

  • *x (tuple[ndarray]) – Input batch.

  • channels (Iterable[str]) – A list of channels to pick.

  • index (int | Iterable[int]) – The index (or indices) of the input ndarray[s] to pick channels from. Default is -1.

  • axis (int) – The channels axis of the input batch. Default is -2.

Returns:

  • *ndarray – Sliced input batch containing only the picked channels.

  • _metadata (dict) – Updated metadata dictionary.