import { PolynomialFeatures, PolynomialFeaturesConfig } from './polynomial-features';
/**
 * For internal use only
 */
export declare type PolynomialRegressorConfig = {
    weights: number[][];
    polyFeatures: PolynomialFeaturesConfig;
};
/**
 * Model for performing multivariate polynomial regression.
 *
 * Train the model by *fit()*ing it on data available for training. Afterwards
 * *predict()*ion is possible.
 *
 * The model is implemented as a pipe consisting of two steps. First the input
 * is transformed by the class PolynomialFeatures, this reduces the problem to a
 * *linear* regression problem. Hence in the second step we simply apply linear
 * regression.
 */
export declare class PolynomialRegressor {
    private _polyFeatures;
    private _weights;
    /**
     * Basic configuration. You can skip configuration by providing no arguments.
     *
     * In case no arguments are given you have to use the method fromConfig(...)
     * to set up a configuration in order to use the method fit(...).
     *
     * For the exact meaning of the arguments see the documentation of
     * PolynomialFeatures.
     *
     * @param degree Highest order of the monomials
     * @param homogeneous Whether to include only highest order monomials
     * @param interactionOnly Whether to disallow higher powers of single features
     */
    constructor(degree?: number, homogeneous?: boolean, interactionOnly?: boolean);
    /** Number of input features. */
    get nFeaturesIn(): number;
    /** Number of output features. */
    get nFeaturesOut(): number;
    /** The weight matrix of the underlying linear regression model. */
    get weights(): number[][];
    /** Instance of PolynomialFeatures responsible for transforming the input. */
    get polyFeatures(): PolynomialFeatures;
    /**
     * Trains the model.
     *
     * @param x List of input vectors.
     * @param y List of corresponding desired output vectors.
     */
    fit(x: number[][], y: number[][]): void;
    /**
     * Make predictions.
     *
     * @param x List of inputs.
     */
    predict(x: number[][]): number[][];
    /**
     * Saves configuration to a simple option-bag
     *
     * The configuration specifies the internal state of a PolynomialRegressor
     * completely. Hence the config of a fitted model can be used to save the
     * model to a file.
    */
    get config(): PolynomialRegressorConfig;
    /**
     * Loads configuration from simple option-bag.
     *
     * @param config The configuration to load from.
     */
    fromConfig(config: PolynomialRegressorConfig): void;
    private predictPoly;
}