Finished up the EM algorithm today. Key was getting the right function to maximize. Turns out wikipedia has a very nice write up of this very example, but in our notation:
\[ E_p \log(p f_1) + (1-E_p) \log( (1-p) f_2 )\]
Where \(E_p\) comes from the expectation step.
Jamie has joined us and has set up a github repository for the discussion group. Find the successful abstract algorithm there.