bobflagg / gibbs-sampling-for-the-uninitiated Goto Github PK

3 questions:

1. for update function in gibbs sampling, can you explain this expression?
   elif C[1] == 1 <= 0:
   in both python 2 and 3, this will always get False, no matter what C[1] is.

2. since you already deduct the document j before:
   C[l] -= 1
   why once again minus 1 here ?
   d = np.sum(C) - 1
v0 = np.log((C[0] - 1.0) / d)
v1 = np.log((C[1] - 1.0) / d)
   this is not consistent with the equation (49) of the paper, right?

3. m is obviously redundant here
   m = max(v0, v1)
v0 = np.exp(v0 - m)
v1 = np.exp(v1 - m)
pi = v1 / (v0 + v1)