num::estimate_largest_eigenvalue

Defined in header <linear/solvers/chebyshev.hpp>

template <class Op>
real estimate_largest_eigenvalue(const Op &A, idx iterations = 25, real safety = 1.1);

Estimate \(\lambda_{\max}\) of a self-adjoint operator by power iteration.

The starting vector is fixed, so repeated calls agree. Power iteration approaches \(\lambda_{\max}\) from below, so the result is multiplied by safety, 10% headroom by default. Returns 0 for an empty operator.