MMLD

With MMLD, the algorithm in subsection has to be used. As the prior function for the binomial distribution is already uniform, equation () can be further simplified.

$$\begin{split} -\log_e f(x\vert\theta) &= 1 - \frac{1}{\int_R... ...nt_R d\theta} \int_R \log_e f(x\vert\theta) d\theta \end{split}$$ (56)

The algorithm to converge to an optimal coding region has to be modified similarly. However, the details of parameter point-estimation for the binomial distribution was not specified previously. It can be shown [#!Dowe:private!#] that -

$$\begin{split} \hat{\theta} &= \frac{1}{\int_R g(\theta\vert m... ...d\theta}\int_R \theta^m (1-\theta)^n \theta d\theta \end{split}$$ (57)