mne_rt.tools.ORICA#

class mne_rt.tools.ORICA(n_channels: int, learning_rate: float = 0.1, block_size: int = 256, online_whitening: bool = True, calibrate_pca: bool = False, forgetfac: float = 1.0, nonlinearity: str = 'tanh', random_state: int | None = None)[source]#

Bases: object

Online Recursive ICA (ORICA) for EEG data.

Parameters:

n_channelsint: Number of input EEG channels.
learning_ratefloat: Learning rate for ICA updates.
block_sizeint: Size of blocks for online updates.
online_whiteningbool: If True, perform online whitening. If False, assumes input is already whitened.
calibrate_pcabool: If True, estimate whitening matrix from the first block and fix it. If False, update recursively.
forgetfacfloat: Forgetting factor for online covariance (0 < forgetfac <= 1). Values < 1 allow adaptation to nonstationary signals.
nonlinearitystr: Nonlinearity: "tanh", "pow3", or "gauss".
random_stateint | None: Seed for reproducibility.

__init__(n_channels: int, learning_rate: float = 0.1, block_size: int = 256, online_whitening: bool = True, calibrate_pca: bool = False, forgetfac: float = 1.0, nonlinearity: str = 'tanh', random_state: int | None = None) → None[source]#

Methods

`__init__`(n_channels[, learning_rate, ...])
`denoise`(X, artifact_idx)	Remove specific ICs and reconstruct sensor data.
`find_blink_ic`(template_map[, threshold])	Identify ICs corresponding to blinks by template correlation.
`fit_transform`(X)	Fit on X and return estimated sources.
`inverse_transform`(S)	Reconstruct sensor-space data from source signals.
`partial_fit`(X)	Update the unmixing matrix with a new block of EEG data.
`transform`(X)	Project data to source space without updating W.
`update_and_denoise`(X, template_map[, threshold])	Fit one block, detect blinks by template, and return cleaned data.

partial_fit(X: ndarray) → ORICA[source]#

Update the unmixing matrix with a new block of EEG data.

Parameters:

Xndarray, shape (n_channels, n_samples): EEG data block.

Returns:

self

transform(X: ndarray) → ndarray[source]#

Project data to source space without updating W.

Parameters:

Xndarray, shape (n_channels, n_samples)

Returns:

Sndarray, shape (n_channels, n_samples)

fit_transform(X: ndarray) → ndarray[source]#

Fit on X and return estimated sources.

Equivalent to partial_fit(X) followed by transform(X).

inverse_transform(S: ndarray) → ndarray[source]#

Reconstruct sensor-space data from source signals.

Applies the full inverse pipeline: X_rec = whitening_inv_ @ pinv(W) @ S + mean_

Parameters:

Sndarray, shape (n_channels, n_samples): Source signals to reconstruct.

Returns:

X_recndarray, shape (n_channels, n_samples)

find_blink_ic(template_map: ndarray, threshold: float = 0.7) → tuple[list[int], ndarray][source]#

Identify ICs corresponding to blinks by template correlation.

Parameters:

template_mapndarray, shape (n_channels,): Spatial topography of a typical blink.
thresholdfloat: Absolute-correlation threshold.

Returns:

blink_idxlist of int
corrsndarray, shape (n_channels,)

denoise(X: ndarray, artifact_idx: list[int]) → ndarray[source]#

Remove specific ICs and reconstruct sensor data.

Parameters:

Xndarray, shape (n_channels, n_samples): Raw EEG data.
artifact_idxlist of int: Component indices to suppress.

Returns:

X_cleanndarray, shape (n_channels, n_samples)

update_and_denoise(X: ndarray, template_map: ndarray, threshold: float = 0.7) → ndarray[source]#

Fit one block, detect blinks by template, and return cleaned data.

Parameters:

Xndarray, shape (n_channels, n_samples)
template_mapndarray, shape (n_channels,)
thresholdfloat

Returns:

X_cleanndarray, shape (n_channels, n_samples)