GaussianMix - Covariance Matrix not positive definite
Hi,
when running the GaussianMixture-PanelGroupings example, sometimes I encounter the error "1. Solution exception:Covariance matrix is not positive definite. Try specifying a regularization constant in the fitting options (there is an example in the Multivariate Normal Distribution documentation)." scrolling through the component number.
I'm currently writing my own version in C# as well before coming across the compiled LunchBox version and came across the fitting option solution given here: http://accord-framework.net/docs/html/T_Accord_Statistics_Distributions_Multivariate_MultivariateNormalDistribution.htm
I saw in the source code that you have tried a similar option but is commented out. Might you be able to discuss a bit more the process and how to incorporate the fitting option into the script if beneficial?
Thank you,
Tim