Prev: compiling and linking c/c++ and .mat files
Next: Weird missing memory usage for a global variable
From: Crystal on 13 Apr 2010 14:10
I am using gmdistirubtion.fit to make finite mixture models of data with a small number of samples. Regularization is required, otherwise an ill-conditioned covariance matrix is created, and gmdistribution.fit allows for adding a regularization parameter (RegVal) to the diagonal of the covariance matrix. I looked in the literature for ways to estimate the RegVal parameter but found only regularization applied to linear systems of the form Ax=b for regression. Can anyone point to literature that describes how to estimate RegVal for a covariance matrix? Or how to rephrase the covariance problem to the form Ax=b so I can use the methods in regression analysis?
 | 
Pages: 1
Prev: compiling and linking c/c++ and .mat files
Next: Weird missing memory usage for a global variable