import { ClassifierBase } from '../base';
import { BaseEstimator, Params } from '../base/estimator';
import { ClassifierLike } from './common';
export interface OneVsRestClassifierProps {
    /** Prototype classifier; one unfitted clone is fitted per class. */
    estimator: BaseEstimator;
}
/**
 * One-vs-rest (one-vs-all) multiclass strategy, mirroring sklearn's
 * `OneVsRestClassifier`.
 *
 * For each of the K classes, a clone of `estimator` is fitted on a binary
 * problem where samples of that class are relabeled 1 and every other sample
 * 0. This turns any binary-only classifier into a multiclass classifier.
 *
 * Prediction picks the class whose member produces the highest score
 * (first class wins ties). The per-member score is, in order of preference:
 * `predictProba`'s positive-class column, `decisionFunction`, or the raw
 * 0/1 `predict` output (vote).
 *
 * Deviations from sklearn (documented):
 *  - sklearn fits a single estimator when y is binary (LabelBinarizer emits
 *    one column); this implementation always fits one member per class.
 *  - `predictProba` rows whose positive-class probabilities sum to 0 fall
 *    back to a uniform distribution instead of producing NaN.
 *
 * Nested params are addressable as `estimator__<param>` in `setParams`.
 */
export declare class OneVsRestClassifier extends ClassifierBase {
    private estimator;
    private classes_;
    private estimators_;
    constructor(props: OneVsRestClassifierProps);
    getParams(): Params;
    /** Supports own params plus nested `estimator__param` addressing. */
    setParams(params: Params): this;
    /** Sorted class labels seen during fit. */
    get classes(): number[];
    /** The fitted per-class members, ordered like `classes`. */
    get estimators(): ClassifierLike[];
    fit(X: number[][], y: number[], sampleWeight?: number[]): void;
    private ensureFitted;
    /** Per-class score matrix [nSamples][nClasses] used by predict. */
    private memberScores;
    predict(X: number[][]): number[];
    /**
     * Positive-class probability of each member, normalized so every row
     * sums to 1 (sklearn behavior). Requires the members to implement
     * `predictProba`.
     */
    predictProba(X: number[][]): number[][];
    /**
     * Per-class decision values, shape [nSamples][nClasses]. Requires the
     * members to implement `decisionFunction`.
     */
    decisionFunction(X: number[][]): number[][];
}