|
What is the covariance matrix Sigma equal to? How do I compute it? |
|
The Sigma for each class is the covariance matrix of the points assigned to it. Usually these are either estimated using the maximum likelihood estimator ( Sigma[i,j] = E[(x[i]-E[x[i]])(x[j]-E[x[j]])] ), but you might want to use the shrinkage estimator instead, Sigma' = (1-lambda) Sigma + lambda I, where Sigma is estimated as above, I is the identity matrix and lambda is a regularization constant. |
