ott.tools.gaussian_mixture.fit_gmm.initialize

Contents

ott.tools.gaussian_mixture.fit_gmm.initialize#

ott.tools.gaussian_mixture.fit_gmm.initialize(rng, points, point_weights, n_components, n_attempts=50, verbose=False)[source]#

Initialize a GMM via K-means++ with retries on failure.

Parameters:

  • rng (Array) – jax.random key

  • points (Array) – (n, n_dimensions) array of observations

  • point_weights (Optional[Array]) – (n,) array of weights for points

  • n_components (int) – desired number of components

  • n_attempts (int) – number of attempts to initialize before failing

  • verbose (bool) – if True, print status information

Return type:

GaussianMixture

Returns:

An initial Gaussian mixture model.

Raises:

ValueError if initialization was unsuccessful after n_attempts attempts. –

Contents