spectral operator computation very slow in scipy 1.14 #14
Description
When using scipy 1.14, the eigendecomposition is getting incredibly slow.
It's happening in
evals_np, evecs_np = sla.eigsh(L_eigsh, k=k_eig, M=Mmat, sigma=eigs_sigma)
in diffusionnet's geometry module