import { ClassifierBase } from '../base';
import { Params } from '../base/estimator';
export interface AdaBoostClassifierProps {
    nEstimators?: number;
    n_estimators?: number;
    learningRate?: number;
    learning_rate?: number;
}
/**
 * AdaBoost classifier.
 *
 * Binary (K = 2): discrete AdaBoost over polarity stumps with the SAMME
 * one-sided weight update.
 *
 * Multiclass (K > 2): SAMME (Zhu et al. 2009), sklearn's AdaBoostClassifier
 * algorithm. The weak learner is a weighted multiclass stump (each side
 * predicts its weighted-majority class), a stump is accepted while its
 * weighted error stays below 1 - 1/K, and
 * alpha = learning_rate * (log((1-err)/err) + log(K-1)).
 */
export declare class AdaBoostClassifier extends ClassifierBase {
    private nEstimators;
    private learningRate;
    private estimators;
    private estimatorWeights;
    private multiEstimators;
    private multiWeights;
    private classes;
    private nFeatures;
    constructor(props?: AdaBoostClassifierProps);
    getParams(): Params;
    private validateInput;
    private isFitted;
    /**
     * trainY here is already mapped to {-1, +1}.
     */
    private trainStump;
    /**
     * Raw stump output in {-1, +1}.
     */
    private stumpScore;
    /**
     * Weighted-misclassification-optimal multiclass stump: for every
     * midpoint threshold each side predicts its weighted-majority class.
     * yIdx holds class indices; weights are assumed normalized to sum 1.
     */
    private trainMultiStump;
    /**
     * Class-index predictions of a multiclass stump.
     */
    private multiStumpPredict;
    fit(trainX: number[][], trainY: number[]): void;
    private fitBinary;
    private fitMulticlass;
    private decisionScores;
    /**
     * Per-class alpha-weighted vote totals, indexed like this.classes.
     */
    private voteScores;
    predict(testX: number[][]): number[];
    /**
     * Returns per-sample probabilities ordered by the sorted class labels.
     * Binary uses a sigmoid over the ensemble score, multiclass the
     * normalized alpha-weighted vote share - both heuristic calibrations,
     * not sklearn's exact probability transform.
     */
    predictProba(testX: number[][]): number[][];
    getFeatureImportances(): number[];
    get featureImportances(): number[];
}