Timeline for Expectation Maximization and Deep Learning
Current License: CC BY-SA 4.0
3 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 6, 2020 at 16:37 | comment | added | Andrew M | @lordanis, many models that apply EM don't require an explicit calculation of the gradient of the expected complete likelihood because the maximization is done exactly and analytically. For instance, the Gaussian finite mixture model. I don't understand what you mean by a "gradient of the expectation step." | |
| Feb 4, 2020 at 22:42 | comment | added | iordanis | I don't understand why you draw a distinction between EM and Gradient Descent. They seem to me 2 different things. We can take the gradient of the Expectation step and for the maximization (or minimization depending on the sign) we can consider the gradient since we can only do a batch at a time. | |
| Feb 4, 2020 at 21:44 | history | answered | Andrew M | CC BY-SA 4.0 |