scot.external package

Submodules

scot.external.infomax_ module

scot.external.infomax_.infomax(data, weights=None, l_rate=None, block=None, w_change=1e-12, anneal_deg=60.0, anneal_step=0.9, extended=False, n_subgauss=1, kurt_size=6000, ext_blocks=1, max_iter=200, random_state=None, verbose=None)

Run the (extended) Infomax ICA decomposition on raw data

based on the publications of Bell & Sejnowski 1995 (Infomax) and Lee, Girolami & Sejnowski, 1999 (extended Infomax)

Parameters:
datanp.ndarray, shape (n_samples, n_features)

The data to unmix.

w_initnp.ndarray, shape (n_features, n_features)

The initialized unmixing matrix. Defaults to None. If None, the identity matrix is used.

l_ratefloat

This quantity indicates the relative size of the change in weights. Note. Smaller learining rates will slow down the procedure. Defaults to 0.010d / alog(n_features ^ 2.0)

blockint

The block size of randomly chosen data segment. Defaults to floor(sqrt(n_times / 3d))

w_changefloat

The change at which to stop iteration. Defaults to 1e-12.

anneal_degfloat

The angle at which (in degree) the learning rate will be reduced. Defaults to 60.0

anneal_stepfloat

The factor by which the learning rate will be reduced once anneal_deg is exceeded: l_rate *= anneal_step Defaults to 0.9

extendedbool

Wheather to use the extended infomax algorithm or not. Defaults to True.

n_subgaussint

The number of subgaussian components. Only considered for extended Infomax.

kurt_sizeint

The window size for kurtosis estimation. Only considered for extended Infomax.

ext_blocksint

The number of blocks after which to recompute Kurtosis. Only considered for extended Infomax.

max_iterint

The maximum number of iterations. Defaults to 200.

verbosebool, str, int, or None

If not None, override default verbose level (see mne.verbose).

Returns:
unmixing_matrixnp.ndarray of float, shape (n_features, n_features)

The linear unmixing operator.

class scot.external.infomax_.logger

Bases: object

static info(*args, **kwargs)

Module contents

External sources and code snippets