mne_denoise.dss.variants.time_shift_dss#

mne_denoise.dss.variants.time_shift_dss(shifts: int | ndarray = 10, *, method: str = 'autocorrelation', n_components: int | None = None, **dss_kws) → DSS[source]#

Create a DSS configured for temporal predictability.

Returns a pre-configured DSS object that extracts components with high autocorrelation (temporally smooth or predictable signals).

Parameters:

  • shifts (int or array-like) – If int, use lags from 1 to shifts. If array, use specified lag values in samples. Default 10.

  • method (str) – Method for constructing bias: - ‘autocorrelation’: Average of shifted versions (default) - ‘prediction’: Weighted average (closer lags weighted more)

  • n_components (int, optional) – Number of DSS components to keep. If None, keep all.

  • **dss_kws – Additional keyword arguments passed to DSS.

Returns:

dss – A DSS object configured with a TimeShiftBias.

Return type:

DSS

Examples

>>> # Extract temporally predictable components
>>> dss = time_shift_dss(shifts=20)
>>> dss.fit(data)
>>> slow_sources = dss.transform(data)

>>> # Use specific lags
>>> dss = time_shift_dss(shifts=np.array([1, 2, 5, 10, 20]))
>>> dss.fit(data)