GaussianMixtures.jl
GaussianMixtures.jl copied to clipboard
EM with missing data
Is there any support for estimating GM model with missing data? There are some algorithms that allow for that (please see attached). EM_Algorithm_MissingData.pdf
This would be currently out of the scope for this package, maybe you can deal with incidental cases in a proprocessing step. I didn't read the full article, though, to really understand the problem.
I implement an EM algorithm with missing data support in BetaML (I didn't follow however the Delalleau paper, so it may not be very efficient with large datasets).
Cool!