[Topic-models] Doubt regarding CORR-LDA
veenat2005 at gmail.com
Wed Nov 5 23:24:03 EST 2008
Thanks Sinead and Jiacheng Guo for clarifying my previous doubt on LDA.
Now I have doubt as to my understanding of the CORR-LDA model as for as the
number of parameters to be estimated.
1. We will have to estimate K number of mean vectors, where each of the
vector is d-dimensional assuming that the feature vector is d-dimensional.
2. Similarly K number of Covariances
3. K number of alphas
4. Coming to beta, is it K number of betas where each beta in turn is V
dimensioal, I mean KxV? Or beta to be estimated is MxV in number?
Please tell me is there any wrong understanding which I have done.
Speech and Vision Lab
IIT Madras, Chennai-36
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Topic-models