scib_metrics.utils.KMeans

class scib_metrics.utils.KMeans(n_clusters=8, init='k-means++', n_init=1, max_iter=300, tol=0.0001, seed=0)

Jax implementation of sklearn.cluster.KMeans.

This implementation is limited to Euclidean distance.

Parameters:

• n_clusters (int (default: 8)) – Number of clusters.

• init (Literal['k-means++', 'random'] (default: 'k-means++')) –

  Cluster centroid initialization method. One of the following:

  • 'k-means++': Sample initial cluster centroids based on an

    empirical distribution of the points’ contributions to the overall inertia.

  • 'random': Uniformly sample observations as initial centroids

• n_init (int (default: 1)) – Number of times the k-means algorithm will be initialized.

• max_iter (int (default: 300)) – Maximum number of iterations of the k-means algorithm for a single run.

• tol (float (default: 0.0001)) – Relative tolerance with regards to inertia to declare convergence.

• seed (Union[int, Array] (default: 0)) – Random seed.

Methods table

fit(X)

Fit the model to the data.

Methods

KMeans.fit(X)

Fit the model to the data.