/** Options for `logisticRegression`. */
export interface LogisticRegressionOptions {
    /** Prepend a column of ones (default true). */
    intercept?: boolean;
    /** Convergence tolerance on the max-norm of the Newton step (default 1e-8). */
    tol?: number;
    /** Maximum IRLS iterations (default 100). */
    maxIter?: number;
}
/** Result of `logisticRegression`. */
export interface LogisticRegressionResult {
    /** Fitted coefficients for the original predictors (excludes the intercept). */
    coefficients: number[];
    /** Fitted intercept (0 if `opts.intercept === false`). */
    intercept: number;
    /** Predict class probabilities (P(y=1|x)) for new rows. */
    predictProba: (x: number[][]) => number[];
    /** Predict class labels (threshold 0.5) for new rows. */
    predict: (x: number[][]) => number[];
}
/**
 * Binary logistic regression `P(y=1|x) = sigmoid(xᵀβ)` fit by IRLS (Newton-Raphson
 * on the Bernoulli log-likelihood).
 *
 * @param X - Design matrix (rows = observations, cols = predictors)
 * @param y - Binary labels, each in {0, 1}
 * @param opts - `intercept` (default true) prepends a column of ones to X;
 *   `tol` (default 1e-8) convergence tolerance; `maxIter` (default 100)
 * @returns Fitted coefficients/intercept plus `predict`/`predictProba`
 *
 * @example
 * const m = logisticRegression([[-2], [-1], [1], [2]], [0, 0, 1, 1]);
 * m.predict([[3]]) // => [1]
 */
export declare function logisticRegression(X: number[][], y: number[], opts?: LogisticRegressionOptions): LogisticRegressionResult;
//# sourceMappingURL=logistic-regression.d.ts.map