From: AJAXDWR@gmail.com   
      
   I just started to read this paper and got confused by the generative   
   process of LDA.   
      
   1. Choose \theta = Dir(\alpha)   
    This might be the only part I understand. \theta is a length-K   
   vector(distribution)   
   if there are K predefined topics.   
      
   2. for each of the N words w_n:   
   (a)choose a topic z_n = mult(\theta)   
   What does this mean? The probability of word w_n to   
   topic z_n? If it is a multinomial distribution, there should   
   be a count x_k besides P(x_k) = \theta_k. Otherwise,   
   how can I calculate this multinomial probability?   
      
   (b) choose a word w_n from p(w_n|z_n,\beta)   
      This is the most confusing part. Since the Step 2 is "for each word   
   w_n",   
    why are we "choose a word w_n" here again? I cannot understand this.   
   Or is this the conditional probability of w_n given z_n and   
   parameters?   
      
   I'm confused by the generative process, how do you actually "generate"   
   words   
   in real application, aren't they contained in the document? For   
   example, if given   
   the following training set   
      
               w_1  w_2   w_3   w_4   
       d1      1     0        3       5   
       d2      0     3        1       2   
   .............................................   
      
   what does the generative process look like? Anyone can help give a   
   walkthrough example?   
   Great thanks.   
      
   [ comp.ai is moderated ... your article may take a while to appear. ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)