R3 3e: a priori information (Remark 2) with heat kernel filters

[iglesias]

The filters are heat kernels given by (before normalization) exp(-tau_i\lambda), with tau_i = 0.3rand+0.7 and \lambda = [1 2 ... L]

  % case DataDistribution.HeatKernel
  h = exp(-(0.3*rand(R, 1)+0.7)*(1:L))';
  % Normalize filter taps.
  h = h ./ repmat(norms(h, 2, 1), L, 1);

[iglesias]

alpha_3 / mu_max with heat kernels more favorable than in other cases?

Done, explained in the response.

[iglesias]

See mu_h_02102022 in github.com/iglesias/gsp_bss

See evaluate_multigraph_bss_direct_18092022 in github.com/iglesias/gsp_bss.