Timeline for Is there any way to derive the initial accuracy with random initialization?
Current License: CC BY-SA 3.0
3 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Dec 28, 2017 at 17:33 | comment | added | Miguel | If you are interested in EM in particular, then this paper might be of interest: "In each case, our theory guarantees that with a suitable initialization, a relatively small number of EM (or gradient EM) steps will yield (with high probability) an estimate that is within statistical error of the MLE." | |
| Dec 28, 2017 at 16:54 | comment | added | user3086871 | is it possible to state how much error reduction is possible after the very first EM iteration for PPCA? Or in general is it possible to give any pessimistic bound on the error after one iteration if started with a random initialization point? | |
| Dec 27, 2017 at 20:48 | history | answered | Miguel | CC BY-SA 3.0 |