I was going through the course again, and I was wondering why the covariance matrix is positive semi-definite and not positive definite.
Wouldn't there be problems if one or more of the eigenvalues were zero? Especially for inversion?
I don't know to be honest, but it does not seem like a mistake as I've searched for it online and the same conclusion about the covariance matrix seem to pop up.
Any student that knows ?