26.10.2013 Views

Nonparametric Bayesian Discrete Latent Variable Models for ...

Nonparametric Bayesian Discrete Latent Variable Models for ...

Nonparametric Bayesian Discrete Latent Variable Models for ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Using the Dirichlet integral,<br />

= α<br />

K<br />

N<br />

i=0<br />

<br />

N<br />

(1 − µ<br />

i<br />

(k))<br />

A.2 Probability of a part of Z being inactive<br />

α<br />

i Γ(i + K<br />

Γ(N + α<br />

K<br />

i=1<br />

)Γ(N − i + 1)<br />

+ 1)<br />

= α<br />

N N!<br />

K (N − i)! i!<br />

i=0<br />

(1 − µ (k)) i (N − i)! i−1 j=0 α<br />

K + j<br />

N j=0 α<br />

K + j<br />

N!<br />

= N j=1 α<br />

<br />

1 +<br />

K + j<br />

α<br />

N<br />

(1 − µ<br />

K<br />

(k)) i<br />

i−1 j=1 α<br />

K + j <br />

.<br />

i!<br />

Raising the above equation to the power K − k and taking the limit as K → ∞,<br />

p(Z (:,.>k) = 0 | µ (k)) = exp − αHN + α<br />

N<br />

i=1<br />

(1 − µ (k)) i<br />

i<br />

(A.9)<br />

. (A.10)<br />

The first term is obtained by using the limit given in eq. (B.14), and the second term is<br />

obtained by the fact that the product inside the sum in the last line of eq. (A.9) reduces<br />

to (i − 1)! in the limit.<br />

Using the series expansion <strong>for</strong> the natural logarithm given in eq. (B.16), the second<br />

term in the above equation can be approximated with a logarithm <strong>for</strong> large N,<br />

p(Z (:,.>k) = 0 | µ (k)) = exp − αHN − α ln(1 − (1 − µ (k))) <br />

=µ −α<br />

(k) exp <br />

− αHN .<br />

(A.11)<br />

117

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!