According to the paper "Bayesian Face Revisited: A Joint Formulation", the repository realizes the algorithm of Joint Beyesian with Matlab and achieve the same result as the paper
matrix A and G should be nonnegative positive semi-definite.
I used zero-centered feature(not the face data) to train, but I got A and G which are not positive semi-definite.
what could be the problem?
matrix A and G should be nonnegative positive semi-definite. I used zero-centered feature(not the face data) to train, but I got A and G which are not positive semi-definite. what could be the problem?