mojolearn.LocallyLinearEmbedding#

class mojolearn.LocallyLinearEmbedding(*, n_neighbors=5, n_components=2, reg=0.001, eigen_solver='auto', tol=1e-06, max_iter=100, method='standard', hessian_tol=0.0001, modified_tol=1e-12, neighbors_algorithm='auto', random_state=None, n_jobs=None, numeric_mode=None)[source]#

Bases: _Base

Locally linear embedding.

Follows scikit-learn’s LocallyLinearEmbedding: reconstructs each point from its Euclidean nearest neighbors and finds the low-dimensional embedding that best preserves those reconstruction weights. Supports the standard, LTSA, Hessian and modified variants.

Parameters:

  • n_neighbors (int, default 5) – Neighbors per point; must be less than n_samples.

  • n_components (int, default 2) – Embedding dimension.

  • reg (float, default 1e-3) – Regularization of the local weight solves.

  • eigen_solver ({“auto”, “dense”}, default “auto”) – "dense" always runs the full exact solve. "auto" with method="standard", more than 200 samples and n_components + 1 < 10 uses an iterative solver for the smallest eigenvectors; otherwise it runs the full solve. "arpack" raises NotImplementedError.

  • tol (ignored)

  • max_iter (int, default 100) – Iteration limit of the iterative solver.

  • method ({“standard”, “ltsa”, “hessian”, “modified”}, default “standard”) – LLE variant. "hessian" needs n_neighbors > n_components * (n_components + 3) / 2; "modified" needs n_neighbors >= n_components.

  • hessian_tol (float, default 1e-4) – Tolerance for method="hessian".

  • modified_tol (float, default 1e-12) – Tolerance for method="modified".

  • neighbors_algorithm (ignored)

  • random_state (int or None, default None) – Seed of the iterative solver’s start; None means 0.

  • n_jobs (ignored)

  • 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:

  • embedding_ (ndarray of shape (n_samples, n_components))

  • reconstruction_error_ (float)

  • n_features_in_ (int)

Raises:

ValueError – For an unknown method or eigen_solver, or n_neighbors out of range for the method.

Examples

>>> import numpy as np, mojolearn
>>> X = np.random.default_rng(0).random((500, 8), dtype=np.float32)
>>> Y = mojolearn.LocallyLinearEmbedding(n_neighbors=10, n_components=2).fit_transform(X)
__init__(*, n_neighbors=5, n_components=2, reg=0.001, eigen_solver='auto', tol=1e-06, max_iter=100, method='standard', hessian_tol=0.0001, modified_tol=1e-12, neighbors_algorithm='auto', random_state=None, n_jobs=None, numeric_mode=None)[source]#

Methods

fit(X[, y])

Compute the embedding of X.

fit_transform(X[, y])

Compute the embedding and return embedding_.

get_params([deep])

Return this estimator's constructor parameters.

set_params(**params)

Set constructor parameters by keyword.

transform(X)

Embed new points.
fit(X, y=None)[source]#

Compute the embedding of X.

Parameters:

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

  • y (ignored)

Returns:

self

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

Compute the embedding and return embedding_.

Parameters:

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

  • y (ignored)

Returns:

ndarray of shape (n_samples, n_components), float32

transform(X)[source]#

Embed new points.

Each query is written as a weighted combination of its nearest training points, and the same weights are applied to their embeddings.

Parameters:

X (array-like of shape (n_queries, n_features))

Returns:

ndarray of shape (n_queries, n_components), float32