mne_denoise.dss.TimeShiftDSS#

class mne_denoise.dss.TimeShiftDSS(*, lag_samples: Sequence[int] | None = None, lag_times: Sequence[float] | None = None, sfreq: float | None = None, n_components: int, rank: int, n_select: int | None = None, component_action: str = 'extract', center: bool = False, distortion_control: str | None = None, reg: float = 1e-09, verbose: bool | str | int | None = None)[source]#

Trial-average DSS in a lag-augmented sensor space.

Parameters:

  • lag_samples (sequence of int | None) – Explicit lag grid in samples. It must contain zero and at least one nonzero lag. Positive lags contribute X(t - lag).

  • lag_times (sequence of float | None) – Explicit lag grid in seconds. Every value must lie on the sampling grid. Exactly one lag representation must be provided.

  • sfreq (float | None) – Sampling frequency for array data when lag_times is used. MNE metadata is authoritative and must agree with a supplied value.

  • n_components (int) – Number of lag-space DSS components to fit. Selection is explicit; there is no in-sample automatic selector.

  • rank (int) – Explicit whitening rank in the augmented feature space.

  • n_select (int | None) – Size of the leading component subspace used by score() and by sensor-space retain or subtract. It is required for those operations; extraction itself can leave it unset.

  • component_action ({'extract', 'retain', 'subtract'}) – Component extraction or sensor-space operation. Sensor operations preserve input shape and leave samples outside the common lag support unchanged.

  • center (bool) – If False (default), use source-aligned uncentered second moments. If True, fit one weighted augmented-feature mean and reuse it for every transform. Epoch-wise and transform-batch centering are never performed.

  • distortion_control ({None, 'cca'}) – Optional paper step 7. 'cca' rotates the fitted reproducible subspace to the single variate most correlated with undelayed training data. It returns one component and requires n_select=1 for sensor operations.

  • reg (float) – Relative numerical rank tolerance used by DSS and optional CCA.

  • verbose (bool | str | int | None) – Logging verbosity.

Notes

Array input is (n_channels, n_times, n_epochs). MNE Epochs input is accepted natively. Continuous and Evoked inputs are intentionally rejected by this initial repeated-trial implementation.

Lag augmentation is the TSDSS-specific layer. The fitted decomposition is available as dss_ and is an ordinary DSS configured with AverageBias(axis="epochs").

Component interpretation and parameter choice require held-out and surrogate validation.

__init__(*, lag_samples: Sequence[int] | None = None, lag_times: Sequence[float] | None = None, sfreq: float | None = None, n_components: int, rank: int, n_select: int | None = None, component_action: str = 'extract', center: bool = False, distortion_control: str | None = None, reg: float = 1e-09, verbose: bool | str | int | None = None) → None[source]#

Methods

__init__(*[, lag_samples, lag_times, sfreq, ...])

fit(X[, y, sample_weight])

Fit lag-augmented repeated-trial DSS filters.

fit_transform(X[, y])

Fit to data, then transform it.

get_metadata_routing()

Get metadata routing of this object.

get_params([deep])

Get parameters for this estimator.

score(X[, y, sample_weight])

Score the leading fitted subspace on held-out repeated trials.

set_fit_request(*[, sample_weight])

Configure whether metadata should be requested to be passed to the fit method.

set_output(*[, transform])

Set output container.

set_params(**params)

Set the parameters of this estimator.

set_score_request(*[, sample_weight])

Configure whether metadata should be requested to be passed to the score method.

transform(X)

Extract components or apply the fitted sensor-space operation.