import { OutlierBase } from '../base';
import { Params } from '../base/estimator';
import { KernelType } from './smo';
export interface OneClassSVMProps {
    kernel?: KernelType;
    degree?: number;
    /** 'scale' = 1/(n_features * Var(X)) like sklearn; 'auto' = 1/n_features */
    gamma?: number | 'scale' | 'auto';
    coef0?: number;
    tol?: number;
    /**
     * upper bound on the fraction of training errors (points classified as
     * outliers), lower bound on the fraction of support vectors
     */
    nu?: number;
    /** hard limit on SMO pair updates; -1 (default) = run until convergence */
    maxIter?: number;
}
/**
 * One-class SVM for unsupervised outlier/novelty detection (Schölkopf's
 * formulation, solved with SMO on the libsvm one-class dual).
 *
 * `decisionFunction(X)` returns f(x) = sum_i alpha_i K(sv_i, x) - rho;
 * `predict(X)` follows the sklearn label convention: **+1 = inlier,
 * -1 = outlier** (positive decision value = inlier).
 *
 * NOTE: this differs from this library's `IsolationForest`, whose `predict`
 * returns 1 for outliers and 0 for inliers.
 */
export declare class OneClassSVM extends OutlierBase {
    private kernel;
    private degree;
    private gamma;
    private coef0;
    private tol;
    private nu;
    private maxIter;
    private gammaValue;
    /** support-vector rows (copies of training samples with alpha_i > 0) */
    private svX;
    /** training-set indices of the support vectors */
    private supportIndices;
    /** unsigned dual coefficients alpha_i, aligned with svX */
    private dualCoef;
    /** decision offset: f(x) = sum alpha_i K(sv_i, x) - rho */
    private rho;
    private fitted;
    constructor(props?: OneClassSVMProps);
    getParams(): Params;
    private kernelConfig;
    fit(samplesX: number[][]): void;
    private checkFitted;
    /** signed distance-like score; positive = inlier, negative = outlier */
    decisionFunction(samplesX: number[][]): number[];
    /** sklearn convention: +1 = inlier, -1 = outlier */
    predict(samplesX: number[][]): number[];
    /** sorted training-set indices of the support vectors (sklearn's support_) */
    getSupportVectors(): number[];
}