On lines 10, 13, and 17 of the pseudocode for Generalized EM for LapPLSI pseudocode, we need to test values of the expected complete data log-likelihood (Q-function). However, in the LapPLSI.m file, the Q-function is never computed. Can someone explain why it isn't necessary to compute the Q-function to test the conditions in the aforementioned lines?
Re: Modeling Hidden Topics on Document Manifolds https://github.com/ZJULearning/MatlabFunc/blob/master/TopicModel/LapPLSI.m
On lines 10, 13, and 17 of the pseudocode for Generalized EM for LapPLSI pseudocode, we need to test values of the expected complete data log-likelihood (Q-function). However, in the LapPLSI.m file, the Q-function is never computed. Can someone explain why it isn't necessary to compute the Q-function to test the conditions in the aforementioned lines?