[R] LDA select number of topics

srecko joksimovic sreckojoksimovic at gmail.com
Sun Nov 8 10:28:34 CET 2015


Hi all,

I've seen recently this great post by Nikita Murzintcev
http://rpubs.com/nikita-moor/107657. If I understood correctly, according
to Griffiths (2004) I should select 11 topics? But, it seems that other
metrics suggest quite different number of topics?

I mean, 11 topics is about the right number, however, besides it works
better in my case, how do I know which metric to rely on? That is, if I
want to report this in a paper, can I simply say that I relied on Griffiths
(2004), without explaining why not Arun (2010), for example?

Thanks,
-------------- next part --------------
A non-text attachment was scrubbed...
Name: dda_topics.pdf
Type: application/pdf
Size: 6272 bytes
Desc: not available
URL: <https://stat.ethz.ch/pipermail/r-help/attachments/20151108/e75038d1/attachment.pdf>


More information about the R-help mailing list