/**
 * 1-D Gaussian kernel density estimation (Wave — ML primitives, Phase 3 Task 5).
 *
 * The first nonparametric density estimator in the library. Given samples
 * `s_1..s_n`, estimates the density at query points via a sum of Gaussian
 * "bumps" centered on each sample:
 *
 *   density(x) = (1 / (n * h)) * sum_i phi((x - s_i) / h)
 *
 * where `phi` is the standard normal pdf and `h` is the bandwidth. Bandwidth
 * defaults to Silverman's rule of thumb, which balances bias (too smooth)
 * against variance (too noisy) using the sample spread.
 */
export interface GaussianKDEOptions {
    /** Bandwidth (smoothing parameter). Defaults to Silverman's rule of thumb. */
    bandwidth?: number;
}
/** Result of `gaussianKDE`: a density evaluator and the bandwidth it uses. */
export interface GaussianKDEResult {
    /** Evaluate the estimated density at each of `xs`. */
    evaluate: (xs: number[]) => number[];
    /** The bandwidth actually used (either supplied or Silverman's rule). */
    bandwidth: number;
}
/**
 * 1-D Gaussian kernel density estimation.
 *
 * @param samples - Observed sample values (n >= 2 for the default bandwidth;
 *   a single sample requires an explicit `opts.bandwidth`)
 * @param opts - `bandwidth` (default: Silverman's rule of thumb)
 * @returns `evaluate(xs)` — density at each query point — and the chosen `bandwidth`
 *
 * @example
 * const kde = gaussianKDE([-1, 0, 0, 1]);
 * kde.evaluate([0]); // => density near the sample center (a single peak)
 */
export declare function gaussianKDE(samples: number[], opts?: GaussianKDEOptions): GaussianKDEResult;
//# sourceMappingURL=kde.d.ts.map