How to implement implicit likelihood model in Edward?


#1

Hi all,

Say P(D|W) = N(wx + wx*x - e^x; 0, I)

So given W as model parameters, we can determine the likelihood of data D as x.

But, it seems that tutorial of Edward only support explicit likelihood model such as

x = N(f(w), I)

Has anyone done this before? Thanks in advance!