pyFM.optimize.weights.resolvent_mask

pyFM.optimize.weights.resolvent_mask(evals1, evals2, gamma=0.5)

LBO Mask, Ren et al., SGP 2019.

The standard (lambda_i - mu_j)^2 mask is unbounded: the energy diverges in k even for C = Id, so no choice of w_lap makes the term well posed. Any bounded g(Delta) has the same null space; this uses the complex resolvent (Delta^gamma - i)^-1. Both spectra are divided by the same factor, max(max Lambda_1, max Lambda_2), so the shapes stay comparable.

Parameters:

  • evals1 ((K1,), (K2,) np.ndarray)

  • evals2 ((K1,), (K2,) np.ndarray)

  • gamma (float) – In (0, 1]. Larger narrows the funnel, i.e. a stronger isometric prior.

Returns:

mask – Rows index mesh2, columns index mesh1 – the orientation of C.

Return type:

(K2, K1) np.ndarray