Timeline for Metropolis-Hastings algorithm for logarithmic probability density
Current License: CC BY-SA 4.0
8 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 13, 2022 at 13:11 | vote | accept | spacexyz | ||
| Feb 13, 2022 at 9:25 | comment | added | Xi'an |
Both methods give exactly the same answer, they are more than equivalent. Note also that the min step is unnecessary for the comparison.
|
|
| Feb 13, 2022 at 3:31 | answer | added | microhaus | timeline score: 0 | |
| Feb 13, 2022 at 3:27 | comment | added | spacexyz | @jassis Thanks for the help - I have managed to get it to work. Out of interest, I have written the code two ways. In one instance I exponentiate $log(\alpha)$ and then continue the method in the image in the OP, in a second method I do the whole method in log format, doing as you said above and taking the log of $u$. Out of curiosity, do you know if the two methods should be equivalent? Or if one would be better than the other? I suspect there may be a difference when comparing $u$ to an exponentiated $log(\alpha)$ vs comparing $log(\alpha)$ to $log(u)$ due to different scales? | |
| Feb 13, 2022 at 2:14 | comment | added | jassis | Precisely! I didn't get to the point, but as I like samplers, I found this post umbertopicchini.wordpress.com/2017/12/18/… | |
| Feb 13, 2022 at 1:54 | comment | added | spacexyz | So draw a random number from 0 to 1, and then take the log of it? | |
| Feb 13, 2022 at 1:53 | comment | added | jassis | I think that if you took $log(\alpha)$ then you will have to take $log(u)$ to preserve inequality. | |
| Feb 13, 2022 at 1:33 | history | asked | spacexyz | CC BY-SA 4.0 |