Timeline for Question on how to use EM to estimate parameters of this model
Current License: CC BY-SA 3.0
12 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jan 22, 2015 at 21:13 | history | edited | Xi'an | CC BY-SA 3.0 |
added 2 characters in body
|
| Jan 22, 2015 at 20:30 | comment | added | Luca | The gradient direction would be completely opposite when you differentiate it? So the sign has to be taken into account, right? | |
| Jan 22, 2015 at 11:37 | comment | added | Luca | Sorry to bother again. In the equation saying.."the part that depends on $\beta$ simplifies as..." should there not be a minus sign in front as well? So basically $Q(\beta|\beta_t)$ is proportional to the negative of the expression that is there. Actually, since we are maximizing it should cancel the negative sign but I wanted to make sure that is what was intended. Sorry, I always second guess myself with the algebra as I have not had enough practice for a very long time. | |
| Jan 22, 2015 at 5:09 | comment | added | Xi'an | correction made: I put $1/2$ in front of the whole expression. | |
| Jan 21, 2015 at 22:58 | comment | added | Luca | Should I edit it to have the square root? | |
| Jan 21, 2015 at 20:50 | history | edited | Xi'an | CC BY-SA 3.0 |
added 11 characters in body
|
| Jan 21, 2015 at 17:46 | comment | added | Luca | One very quick thing...I am not sure if you see this but when you have the equation for the complete log-likelihood, is the first term not $log(\sqrt{w_i})$? Also, I am guessing the term you show is the log-likelihood proportional to a constant. I always get confused with this when stuff gets rolled up into constants. | |
| Jan 21, 2015 at 11:27 | vote | accept | Luca | ||
| Jan 20, 2015 at 17:22 | comment | added | Luca | Perhaps what they have dome is the MAP estimate instead of ML estimate. If I try and reformulate this as the MAP estimate, I am guessing the prior distribution of $\beta$ would come into play? | |
| Jan 20, 2015 at 17:13 | comment | added | Xi'an | If you treat both $\beta$ and $w$ as latent variables, there is no parameter left... | |
| Jan 20, 2015 at 17:07 | comment | added | Luca | Thanks for this and I will go through this rigorously. However, this work I am looking at does treat $\beta$ as a hidden variable too. They mention they take the expectation with the approximate form of posterior $Q(\beta, w)$ approximating it as $Q(w)Q(\beta)$. So this bit has me really confused... | |
| Jan 20, 2015 at 16:56 | history | answered | Xi'an | CC BY-SA 3.0 |