16:30:51	 From Dan Liyanage : Why can’t you compute the p(w)?
16:35:38	 From Prasanth Shyamsundar : Is the latent class variable beta discrete, or can it be continuous too?
16:36:39	 From Prasanth Shyamsundar : Ignore this question
16:54:49	 From Prasanth Shyamsundar : Is beta a class label denoting topics, or a continuous variable denoting the weight of a given topic?
17:01:31	 From Dimitri Bourilkov : Can you please show again the while...do slide.
17:02:59	 From Tom Junk : What determines the dimensionality of nu?  In clustering applications, there is almost always a cluster-size parameter which steers the desired granularity.  What keeps the method from creating a different topic class for each paper?
17:05:57	 From Prasanth Shyamsundar : The learned posterior distribution of latent variables in the LDA analysis is different from posteriors in a Bayesian sense (where more data makes the posteriors tighter), right?