Let $\Sigma $ be the covariance matrix of a multivariate normal distribution.
Thereotically, I guess it does not have to to strictly positive definition, i.e., $\det (\Sigma )=0$ is possible.
In this case, the density function does not exist, how can I generate random numbers(vectors) from such multivariate normal distribution?