import { RegressorBase } from '../base';
import { Params } from '../base/estimator';
import { KernelConfig, KernelMatrix, KernelType, SVRSolution } from './smo';
export interface SVRProps {
    kernel?: KernelType;
    degree?: number;
    /** 'scale' = 1/(n_features * Var(X)) like sklearn; 'auto' = 1/n_features */
    gamma?: number | 'scale' | 'auto';
    coef0?: number;
    tol?: number;
    C?: number;
    /** half-width of the epsilon-insensitive tube (no penalty for |f(x)-y| <= epsilon) */
    epsilon?: number;
    /** hard limit on SMO pair updates; -1 (default) = run until convergence */
    maxIter?: number;
}
/**
 * epsilon-Support Vector Regression solved in the dual with SMO on the
 * libsvm 2n-variable formulation (alpha and alpha* pairs, maximal-violating-
 * pair working-set selection). The fitted model is
 * f(x) = sum_k dualCoef_k K(sv_k, x) + intercept with dualCoef = alpha - alpha*.
 */
export declare class SVR extends RegressorBase {
    protected kernel: KernelType;
    protected degree: number;
    protected gamma: number | 'scale' | 'auto';
    protected coef0: number;
    protected tol: number;
    protected C: number;
    protected epsilon: number;
    protected maxIter: number;
    protected gammaValue: number;
    /** support-vector rows (copies of the training samples with nonzero dual coef) */
    protected svX: number[][];
    /** training-set indices of the support vectors */
    protected supportIndices: number[];
    /** signed dual coefficients alpha_k - alpha*_k, aligned with svX */
    protected dualCoef: number[];
    protected intercept: number;
    protected fitted: boolean;
    constructor(props?: SVRProps);
    getParams(): Params;
    protected kernelConfig(): KernelConfig;
    /** solve the regression dual; overridden by NuSVR */
    protected solveDual(K: KernelMatrix, y: number[]): SVRSolution;
    fit(trainX: number[][], trainY: number[]): void;
    protected checkFitted(): void;
    predict(testX: number[][]): number[];
    /** sorted training-set indices of the support vectors (sklearn's support_) */
    getSupportVectors(): number[];
}