mne_denoise.dss.denoisers.KurtosisDenoiser

class mne_denoise.dss.denoisers.KurtosisDenoiser(nonlinearity: str = 'tanh', alpha: float = 1.0)

Kurtosis maximization denoiser.

Can wrap different nonlinearities (‘tanh’, ‘cube’, ‘gauss’) to maximize non-Gaussianity. Included for checking various ICA contrasts.

Parameters:

• nonlinearity ({'tanh', 'cube', 'gauss'}) – The function $g(s)$ to use. ‘cube’ ($s^3$) is the classic kurtosis maximization.

• alpha (float) – Scaling parameter.

Examples

>>> from mne_denoise.dss.denoisers import KurtosisDenoiser
>>> denoiser = KurtosisDenoiser(nonlinearity="cube")
>>> denoised = denoiser.denoise(source)

References

Särelä & Valpola (2005). Section 4.2.1 “KURTOSIS-BASED ICA”

__init__(nonlinearity: str = 'tanh', alpha: float = 1.0) → None

Methods

__init__([nonlinearity, alpha])
denoise(source)	Apply nonlinearity.