# [Topic-models] Dear all, how to estimate a Dirichlet parameter? Thanks.

Xianling Mao mxl.cs.pku at gmail.com
Sat Oct 22 20:29:52 EDT 2011

Dear all,
During developing a topic model, I encountered a problem:
there is a parent node and n sub nodes. For each of sub nodes, we have known
its times that occurred, let them be t_1, t_2, ..., t_n.
We assume that we first choose a multinomial distribution for the parent
node: \theta ~ p(\theta| \alpha_1, ..., \alpha_n),
then we choose a sub node by this multinomial distribution \theta.
Now, I can not estimate the \alpha parameter, anybody can give me some clue?
Thanks!

--
Best Wishes!
--------------------------------------
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/topic-models/attachments/20111023/79f92a08/attachment.htm>