/**
 * Calculus connectors (Wave C / bridge C3).
 *
 * `hessian` is the missing second-order object between calculus, optimization, and
 * linear algebra — second-order optimizers and Laplace approximations need it.
 * Numeric central differences over any `f: ℝⁿ → ℝ` (no AD-compatibility required);
 * complements the symbolic `jacobian`/`gradientSymbolic` and the AD `gradientAt`.
 */
type Vec = readonly number[] | Float64Array;
type ScalarField = (x: number[]) => number;
/**
 * Numeric gradient of sampled data `y` over abscissae `x` (unit spacing, a scalar
 * `dx`, or an explicit grid). Central differences in the interior, one-sided at the
 * edges — matches `numpy.gradient`. Distinct from the AD `gradientAt` (this works on
 * sampled values, not a differentiable function).
 */
export declare function gradient(y: Vec, x?: Vec | number): number[];
/**
 * Numeric Hessian (matrix of second partials) of `f` at `x` via central
 * differences. Symmetric by construction. `h` is the step (default 1e-4, a good
 * tradeoff between truncation and round-off for double precision).
 */
export declare function hessian(f: ScalarField, x: readonly number[], h?: number): number[][];
export {};
//# sourceMappingURL=calculus-extra.d.ts.map