Skip to content

Conversation

@CorcovadoMing
Copy link
Contributor

should shuffle the datasets first in order to get the correct validation result

@wepe
Copy link
Owner

wepe commented May 16, 2015

CorcovadoMing, you are right. Many thanks!
And I think it is not necessary to shuffle the data each epoch, I will remove the same code following.

wepe added a commit that referenced this pull request May 16, 2015
@wepe wepe merged commit 7d68d09 into wepe:master May 16, 2015
@CorcovadoMing
Copy link
Contributor Author

It's hard to say whether shuffle in each epoch needed, but I think maybe it's a good idea to shuffle in each epochs to avoid a fixed training pattern

need more experiments to confirm which one is better :)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants