hyperspy_ml_algorithms.Orthomax

class hyperspy_ml_algorithms.Orthomax(gamma=1.0, tol=0.001, max_iter=100)

Bases: object

Orthomax rotation of components (varimax when gamma=1.0).

Computes an orthogonal rotation of the component matrix, preserving orthogonality of the components. When gamma=1.0 this is the standard varimax rotation, which finds a rotation matrix W that maximises the variance of the squared components.

Parameters:

gammafloat, default=1.0

Orthomax parameter. If gamma in range [0, 1], use the SVD-based fast algorithm, otherwise solve with a sequence of bivariate rotations. The default (1.0) corresponds to varimax.

tolfloat, default=1e-3

Tolerance of the stopping condition.

max_iterint, default=100

Maximum number of iterations before exiting without convergence.

Attributes:

rotation_matrix_ndarray of shape (n_components, n_components)

The learned rotation matrix.

components_ndarray of shape (n_features, n_components)

The rotated components matrix (input X multiplied by rotation_matrix_).

Notes

Orthomax operates on a component matrix, not raw data. The input X must have shape (n_features, n_components); pass the transpose of components produced by sklearn-style decompositions, e.g. Orthomax().fit(pca.components_.T).

Adapted from metpy.

__init__(gamma=1.0, tol=0.001, max_iter=100)

Methods

__init__([gamma, tol, max_iter])
fit(X[, y])	Fit the orthomax rotation to the component matrix X.
fit_transform(X[, y])	Fit the orthomax rotation to X and return the rotated components.
transform(X)	Apply the learned rotation to a component matrix X.