num::gaussian_probe

Defined in header <stochastic/probe.hpp>

template <typename RNG = rng>
mat<real> gaussian_probe(idx n, idx probes, RNG &generator);  // (1)

mat<real> gaussian_probe(idx n, idx probes, unsigned seed);  // (2)

(1) An \(n \times p\) matrix of independent standard Gaussian probes.

Like rademacher_probe, column \(k\) satisfies \(E[z_k z_k^T] = I\), so either family drives a Hutchinson estimator. Gaussian probes are the ones to reach for when the estimator's concentration is stated through a chi-square tail bound, since each row of the probed block is then exactly chi-square.

(2) The same Gaussian probe from a fixed seed.