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Outdated suggestions cannot be applied.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.
This PR adds a max function option to the elementwise layer as suggested in #654 . Tests are included, however, all four gradient tests fail using the elementwise checker!
For the gradient computation I assume that only the gradients of maximum values in forward pass are propagated backwards, while all the remaining values are set to zero. I wonder if the gradient calculation is flawed or the gradient checker needs to be adapted to deal with the elementwise max case. Maybe someone familiar with the gradient checker could have a look.
Here is some output from the GPU test run: