mojolearn.FastICA#

class mojolearn.FastICA(n_components=None, *, algorithm='parallel', whiten='unit-variance', fun='logcosh', fun_args=None, max_iter=200, tol=0.0001, w_init=None, whiten_solver='svd', random_state=None, numeric_mode=None)[source]#

Bases: _Base

Independent component analysis with the FastICA algorithm.

Follows scikit-learn’s FastICA. Whitening always uses the eigendecomposition of X.T @ X (whiten_solver is accepted but both values take this route), and the default starting matrix is drawn from mojolearn’s own generator, so results can differ from scikit-learn’s for the same seed.

Parameters:

  • n_components (int or None, default None) – Number of components; None, or a value above min(n_samples, n_features), uses min(n_samples, n_features). Ignored when whiten=False.

  • algorithm ({“parallel”, “deflation”}, default “parallel”) – Estimate all components together, or one at a time.

  • whiten ({“unit-variance”, “arbitrary-variance”} or False, default “unit-variance”) – Whitening strategy; False assumes X is already whitened.

  • fun ({“logcosh”, “exp”, “cube”}, default “logcosh”) – Contrast function. A callable is not supported.

  • fun_args (dict or None, default None) – For "logcosh", {"alpha": a} with 1 <= a <= 2 (default 1.0).

  • max_iter (int, default 200) – Maximum number of iterations.

  • tol (float, default 1e-4) – Convergence tolerance.

  • w_init (array-like of shape (n_components, n_components) or None, default None) – Starting unmixing matrix; None draws one from random_state.

  • whiten_solver ({“svd”, “eigh”}, default “svd”) – Accepted for scikit-learn compatibility; see above.

  • random_state (int or None, default None) – Seed for the default w_init; None means 0.

  • numeric_mode ({“identical”, “fast”} or None, default None) – Numeric mode for this estimator. None uses the package default ("identical" unless changed). "identical" gives the same bits on NVIDIA and AMD GPUs; "fast" applies on Apple GPUs.

Variables:

  • components_ (ndarray of shape (n_components, n_features)) – Unmixing matrix applied by transform.

  • mixing_ (ndarray of shape (n_features, n_components)) – Pseudo-inverse of components_.

  • mean_ (ndarray of shape (n_features,)) – Only when whitening.

  • whitening_ (ndarray of shape (n_components, n_features)) – Only when whitening.

  • n_iter_ (int)

  • n_features_in_ (int)

Raises:

ValueError – For an unsupported algorithm, whiten or fun, alpha outside [1, 2], or a w_init of the wrong shape.

Examples

>>> import numpy as np, mojolearn
>>> X = np.random.default_rng(0).random((1000, 8), dtype=np.float32)
>>> S = mojolearn.FastICA(n_components=3, random_state=0).fit_transform(X)
__init__(n_components=None, *, algorithm='parallel', whiten='unit-variance', fun='logcosh', fun_args=None, max_iter=200, tol=0.0001, w_init=None, whiten_solver='svd', random_state=None, numeric_mode=None)[source]#

Methods

fit(X[, y])

Fit the model.

fit_transform(X[, y])

Fit the model and return the estimated sources.

get_params([deep])

Return this estimator's constructor parameters.

inverse_transform(X[, copy])

Map sources back to the feature space with mixing_.

set_params(**params)

Set constructor parameters by keyword.

transform(X[, copy])

Recover the sources from X.
fit_transform(X, y=None)[source]#

Fit the model and return the estimated sources.

Parameters:

  • X (array-like of shape (n_samples, n_features))

  • y (ignored)

Returns:

ndarray of shape (n_samples, n_components), float32

fit(X, y=None)[source]#

Fit the model.

Parameters:

  • X (array-like of shape (n_samples, n_features))

  • y (ignored)

Returns:

self

transform(X, copy=True)[source]#

Recover the sources from X.

Parameters:

  • X (array-like of shape (n_samples, n_features))

  • copy (bool, default True) – Accepted for scikit-learn compatibility; X is never modified.

Returns:

ndarray of shape (n_samples, n_components), float32

inverse_transform(X, copy=True)[source]#

Map sources back to the feature space with mixing_.

Parameters:

  • X (array-like of shape (n_samples, n_components))

  • copy (bool, default True) – Accepted for scikit-learn compatibility; X is never modified.

Returns:

ndarray of shape (n_samples, n_features), float32