Implementing the Decayed Adagrad optimizer operator#4645
Merged
kexinzhao merged 4 commits intoPaddlePaddle:developfrom Oct 12, 2017
Merged
Implementing the Decayed Adagrad optimizer operator#4645kexinzhao merged 4 commits intoPaddlePaddle:developfrom
kexinzhao merged 4 commits intoPaddlePaddle:developfrom
Conversation
jacquesqiao
reviewed
Oct 9, 2017
|
|
||
| Decayed Adagrad | ||
|
|
||
| moment_out = decay * moment + (1 - decay) * grad * grad |
Member
There was a problem hiding this comment.
does there have an article about this optimize algorithm?
Contributor
Author
There was a problem hiding this comment.
I cannot find published article about it.
Just following the formula in http://doc.paddlepaddle.org/develop/doc/api/v2/config/optimizer.html#decayedadagrad
23 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes #4643
Formula for decayed Adagrad operator:
moment_out = decay * moment + (1 - decay) * grad * grad
param_out = param - learning_rate * grad / (sqrt(moment_out) + epsilon)