import { ClassifierBase } from '../base';
import { Params } from '../base/estimator';
export interface GradientBoostingClassifierProps {
    nEstimators?: number;
    n_estimators?: number;
    learningRate?: number;
    learning_rate?: number;
    maxDepth?: number;
    max_depth?: number;
    minSamplesSplit?: number;
    min_samples_split?: number;
    subsample?: number;
    maxFeatures?: number | 'sqrt' | 'log2';
    max_features?: number | 'sqrt' | 'log2';
    randomState?: number;
    random_state?: number;
}
/**
 * Gradient boosting classifier following sklearn's GradientBoostingClassifier.
 *
 * Binary (K = 2): log loss. F_0 is the log-odds of the positive class, each
 * round fits one regression tree to the negative gradient y - p and replaces
 * each leaf value with the Newton step sum(residuals) / sum(p * (1 - p)).
 *
 * Multiclass (K > 2): multinomial deviance. F_0k = log(prior_k), each round
 * fits K trees (one per class) to y_k - softmax_k(F) using the shared
 * pre-round probabilities, with the Newton step
 * (K-1)/K * sum(residuals) / sum(p * (1 - p)) per leaf.
 */
export declare class GradientBoostingClassifier extends ClassifierBase {
    private nEstimators;
    private learningRate;
    private maxDepth;
    private minSamplesSplit;
    private subsample;
    private maxFeatures?;
    private randomState?;
    private estimators;
    private multiEstimators;
    private classes;
    private initF;
    private initFMulti;
    private fitted;
    constructor(props?: GradientBoostingClassifierProps);
    getParams(): Params;
    private sampleRows;
    private buildTree;
    /**
     * Fit one tree to the residuals over the in-bag rows and replace its
     * leaf values with the Newton step
     * factor * sum(residuals) / sum(p * (1 - p)).
     */
    private fitNewtonTree;
    fit(trainX: number[][], trainY: number[]): void;
    private fitBinary;
    private fitMulticlass;
    /**
     * Signed distance to the decision boundary: number[] for binary,
     * number[][] (per-class margins) for multiclass.
     */
    decisionFunction(testX: number[][]): number[] | number[][];
    private decisionFunctionBinary;
    private decisionFunctionMulti;
    /**
     * Returns per-sample probabilities ordered by the sorted class labels.
     */
    predictProba(testX: number[][]): number[][];
    predict(testX: number[][]): number[];
    get featureImportances(): number[];
}