[Topic-models] Variance of corpus?

Gregor Heinrich gregor at arbylon.net
Sat May 16 06:43:16 EDT 2009


Hi -- as hyperparameters influence the "natural" number of topics K, you 
need to train them as well to be "objective" about K. Holding them 
constant will give you an optimum K given the chosen hyperparameters.

Best regards

gregor

Aaron Zinman wrote:
> You can always hold your hyperparameters constant and use perplexity 
> or empirical likelihood.
>
> Aaron
>
> On Fri, May 15, 2009 at 6:47 PM, David André Broniatowski 
> <david at mit.edu <mailto:david at mit.edu>> wrote:
>
>     Dear topic model list,
>
>     Please excuse me if this is a very basic question, but I'm
>     wondering if there is a principled Bayesian way to measure the
>     variance of a corpus.
>
>     Ultimately, I'm trying to figure out how many topics to choose in
>     order to fit a topic model to a corpus.
>
>     Thanks,
>     David
>
>     _______________________________________________
>     Topic-models mailing list
>     Topic-models at lists.cs.princeton.edu
>     <mailto:Topic-models at lists.cs.princeton.edu>
>     https://lists.cs.princeton.edu/mailman/listinfo/topic-models
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> Topic-models mailing list
> Topic-models at lists.cs.princeton.edu
> https://lists.cs.princeton.edu/mailman/listinfo/topic-models
>   
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/topic-models/attachments/20090516/3cd3246c/attachment.html>


More information about the Topic-models mailing list