Description

FactorAnalysis.fit isn't using EM algorithm , even though written "EM" in doc?

I can't see calculating hidden variable's posterior dist in E-step and updating parameters using this posterior in M-step.

This implementation isn't just "iterative method based-on SVD" ?

(Because selecting initial values is hard on the follow EM algorithm, I think not wrong this implemented method .)

quote: "Bayesian Reasoning and Machine Learning" p.394, http://web4.cs.ucl.ac.uk/staff/D.Barber/textbook/090310.pdf