import { ClusterBase } from '../base/cluster';
import { Params } from '../base/estimator';
export interface KMeansProps {
    /** number of clusters */
    n_clusters?: number;
    /**
     * relative tolerance on the weighted inertia change used as convergence
     * criterion (previously named `opt_ratio`; old positional calls keep
     * working and map to `tol`)
     */
    tol?: number;
    /**
     * optional user-provided initial centers; when given, they are used
     * as-is and `n_init` is forced to 1
     */
    initCenters?: number[][];
    /** maximum Lloyd iterations per run */
    max_iter?: number;
    /** optional seed for reproducible k-means++ init */
    random_state?: number;
    /**
     * number of k-means++ restarts; the run with the lowest weighted
     * inertia wins
     */
    n_init?: number;
}
export declare class KMeans extends ClusterBase {
    private n_clusters;
    private tol;
    private max_iter;
    private n_init;
    private random_state?;
    private userCenters;
    private centers;
    private samplesY;
    private inertia;
    constructor(props?: KMeansProps);
    /** @deprecated positional form; prefer the props-object constructor */
    constructor(n_clusters?: number, tol?: number, initCenters?: number[][], max_iter?: number, random_state?: number, n_init?: number);
    getParams(): Params;
    private assignLabels;
    private updateCentroids;
    private runSingle;
    fitPredict(sampleX: number[][], sampleWeights?: number[]): number[];
    getCentroids(): number[][] | null;
    getInertia(): number;
}