Closed blegat closed 3 years ago
orthogonal bases are not really possible with exact/symbolic numbers; for now if you specify Float64
(or Complex64
) as the first argument to symmetry_adapted_basis
svd
will be used to find an orthogonal basis, otherwise we keep it as is (see the docstring for symmetry_adapted_basis
).
With the group in https://gist.github.com/kalmarek/37fde1975162b9709f72d6a312fcbf91, the basis is
which gives the induced matrices
and
which are not orthogonal. The SDP solution is
which has eigenvalues
(1, 3)
and eigenvectors(1, 1)
and(1, -1)
so we guess a different basiswhich has orthogonal induced matrices. Would it be possible for SymbolicWedderburn to return this basis instead ?