mojolearn.NearestCentroid#

class mojolearn.NearestCentroid(metric='euclidean', *, shrink_threshold=None, priors='uniform')#

Nearest centroid classifier.

Represents each class by its centroid (the per-class mean for metric='euclidean', the per-class median for 'manhattan') and assigns each sample to the nearest one, following scikit-learn’s NearestCentroid, including shrunken centroids and class priors. With non-uniform priors, predict uses the discriminant of decision_function(). Sparse input is densified. Computation is float32 (scikit-learn uses float64). A feature whose shrink scale is zero gets deviation 0 (scikit-learn divides by zero there).

Parameters:

  • metric ({‘euclidean’, ‘manhattan’}, default ‘euclidean’) – Distance used for centroids and prediction; anything else raises ValueError.

  • shrink_threshold (float or None, default None) – Threshold for shrinking centroids toward the overall centroid; None disables shrinking.

  • priors ({‘uniform’, ‘empirical’} or array-like of shape (n_classes,), default ‘uniform’) – Class priors. 'empirical' uses the class frequencies. Given priors must be non-negative, one per class, and are rescaled to sum to one.

  • numeric_mode ({‘identical’, ‘fast’} or None, default None) – Numeric mode for this estimator ('fast' is Apple only). None uses the process default.

Variables:

  • centroids_ (Array of shape (n_classes, n_features)) – Class centroids, after shrinking.

  • deviations_ (Array of shape (n_classes, n_features)) – Shrunken deviations of each class centroid from the overall centroid.

  • within_class_std_dev_ (Array of shape (n_features,)) – Pooled within-class standard deviation of each feature.

  • class_prior_ (Array of shape (n_classes,), float64) – Class priors.

  • classes_ (list) – Class labels.

  • n_features_in_ (int) – Number of features seen in fit.

Examples

>>> import numpy as np, mojolearn
>>> rng = np.random.default_rng(0)
>>> X = rng.random((1000, 8), dtype=np.float32)
>>> y = (X[:, 0] > 0.5).astype(np.int64)
>>> clf = mojolearn.NearestCentroid().fit(X, y)
>>> pred = clf.predict(X)

See also

User guide: Neighbors and vector search

Choosing an estimator

Methods

decision_function(X)

Discriminant score of each sample for each class.

fit(X, y)

Fit the class centroids.

get_params([deep])

numeric_mode_used()

The numeric mode this estimator runs in.

predict(X)

Predict the class of each sample.

predict_log_proba(X)

Log class probabilities: the log-softmax of decision_function().

predict_proba(X)

Class probabilities: the softmax of decision_function().

score(X, y[, sample_weight])

Mean accuracy of predict() on the given data.

set_params(**params)

Set constructor parameters.

vendor_used()

The GPU backend this estimator runs on.
fit(X, y)#

Fit the class centroids.

Parameters:

  • X (array-like or scipy.sparse matrix of shape (n_samples, n_features)) – Training data.

  • y (array-like of shape (n_samples,)) – Class labels; at least two classes.

Returns:

self (NearestCentroid) – The fitted estimator.

Raises:

ValueError – If there are fewer than two classes, the priors are invalid, or every feature has zero variance.

predict(X)#

Predict the class of each sample.

With uniform priors this is the class of the nearest centroid; otherwise the class with the largest decision_function() value.

Parameters:

X (array-like of shape (n_queries, n_features)) – Samples to classify.

Returns:

labels (Array or list of shape (n_queries,)) – Predicted class labels.

decision_function(X)#

Discriminant score of each sample for each class.

Available for metric='euclidean' only; otherwise raises AttributeError.

Parameters:

X (array-like of shape (n_queries, n_features)) – Samples to score.

Returns:

scores (Array of shape (n_queries, n_classes), float32)

predict_proba(X)#

Class probabilities: the softmax of decision_function().

Available for metric='euclidean' only.

Parameters:

X (array-like of shape (n_queries, n_features)) – Samples to score.

Returns:

proba (Array of shape (n_queries, n_classes)) – Probabilities, columns in the order of classes_.

predict_log_proba(X)#

Log class probabilities: the log-softmax of decision_function().

Available for metric='euclidean' only.

Parameters:

X (array-like of shape (n_queries, n_features)) – Samples to score.

Returns:

log_proba (Array of shape (n_queries, n_classes), float32)

score(X, y, sample_weight=None)#

Mean accuracy of predict() on the given data.

Parameters:

  • X (array-like of shape (n_samples, n_features)) – Test samples.

  • y (array-like of shape (n_samples,)) – True labels.

  • sample_weight (array-like of shape (n_samples,) or None, default None) – Sample weights.

Returns:

score (float) – (Weighted) fraction of correct predictions.

numeric_mode_used()#

The numeric mode this estimator runs in.

Returns:

str – 'identical' or 'fast', read from the compiled binary the estimator uses, not from the argument it was given.

set_params(**params)#

Set constructor parameters. The estimator is rebuilt through its constructor, so every constructor check runs again and any fitted state is dropped, as a new setting invalidates it.

vendor_used()#

The GPU backend this estimator runs on.

Returns:

str – 'metal', 'cuda' or 'hip', read from the compiled binary the estimator uses.