Homepage › Solution manuals › Kevin P. Murphy › Machine Learning: a Probabilistic Perspective › Exercise 12.2 - MAP estimation for the FA model

Machine Learning: a Probabilistic Perspective

Exercise 12.2 - MAP estimation for the FA model

Answers

Assume that the model is exerted the prior distribution $p (𝐖)$ and $p (Ψ)$ . Compare with the question before, the M-step needs to be modified into:

\begin{aligned} \frac{\partial}{\partial 𝐖} (Q + \log p (𝐖)) = 0, \\ \frac{\partial}{\partial Ψ} (Q + \log p (Ψ)) = 0, \end{aligned}

whose results dependent on the concrete form of the prior distributions.

solour_lfq

2021-03-24 13:42

Comments