- If you see oscillating training loss, likely the descent steps are bouncing between two "valleys" - try using MomentumOptimizer.
- Playing around with the step_size can help avoid non-covergence to nan
- The std of the random_normal of the initial weights should be quite small (ex:
stddev=0.01
) - Also using RMSPropOptimizer really seems to help decrease the obj function (at least, better than the GradientDescent)
CamDavidsonPilon/tf-examples
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
About
my tf examples for now
Resources
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published