export interface MakeRegressionProps {
    /** number of samples */
    nSamples?: number;
    /** number of features */
    nFeatures?: number;
    /** number of features actually used to build the linear model */
    nInformative?: number;
    /** number of regression targets */
    nTargets?: number;
    /** bias term of the underlying linear model */
    bias?: number;
    /**
     * approximate number of singular vectors required to explain most of
     * the input variance; when set, X becomes a low-rank matrix with a
     * bell-shaped singular value profile instead of i.i.d. gaussian noise
     */
    effectiveRank?: number;
    /** relative importance of the fat noisy tail of the singular profile (with effectiveRank) */
    tailStrength?: number;
    /** standard deviation of the gaussian noise added to the targets */
    noise?: number;
    /** shuffle the samples and the feature columns (default true) */
    shuffle?: boolean;
    /** also return the ground-truth coefficients */
    coef?: boolean;
    /** seed for reproducible output */
    randomState?: number;
}
export interface MakeRegressionResult {
    /** samples, shape [nSamples][nFeatures] */
    X: number[][];
    /** targets: number[] when nTargets === 1, number[][] ([nSamples][nTargets]) otherwise */
    y: number[] | number[][];
    /**
     * ground-truth coefficients (only when `coef: true`): number[] of
     * length nFeatures when nTargets === 1, number[][] ([nFeatures][nTargets]) otherwise
     */
    coef?: number[] | number[][];
}
/**
 * Generates a random regression problem, y = X . coef + bias + noise
 * (port of sklearn.datasets.make_regression). Only nInformative features
 * have non-zero coefficients (uniform in [0, 100)).
 *
 * `effectiveRank` is fully implemented (sklearn's make_low_rank_matrix):
 * X = U diag(s) V^T with orthonormal U, V from QR of gaussian matrices and
 * a bell-shaped + fat-tail singular value profile controlled by tailStrength.
 */
export declare function makeRegression(props?: MakeRegressionProps): MakeRegressionResult;
/**
 * Mostly low-rank matrix with bell-shaped singular values
 * (port of sklearn.datasets.make_low_rank_matrix).
 */
export declare function makeLowRankMatrix(nSamples: number, nFeatures: number, effectiveRank: number, tailStrength: number, gaussian: () => number): number[][];