According to Wikipedia, the parameter in a Bernoulli distribution should be $0<p<1$.

I am reading this famous paper proposing Hierarchical Dirichlet Process, and on page 1580, A.6 and the sentence right after it, they state


but $n_{j..}$ is an integer which can take any large value, while $\alpha_0$ is a real value which almost always is way below $n_{j..}$. Also $s_j$ is a binary value either 0, or 1.

How can I convert it to an actual Bernoulli?


Since $p(1)=p$ and $p(0)=1-p$ are both proportional to a known expression* (the unscaled probabilities, $u(i)=c.p(i)$, with the same unknown constant of proportionality, $c$) and you know the $p(i)$ values must add to $1$, then $u(0)+u(1)=c$.

Which is to say $p(i) = \frac{u(i)}{u(0)+u(1)},\: i=0,1$.

(This notion is widely used in Bayesian statistics with many kinds of discrete variables.)

Note that $u(0)=1$ (always, since the power is $s_i=0$), so $p(0) = 1/(1+u(1))$ and $p(1)=u(1)/(1+u(1))$

*(these $p$'s are $q(s_i|\alpha_0)$ in the paper, for $s_i=0$ and $1$ respectively)

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.