EXP neuron layer

[jeffdonahue]

Computes y = base ^ (scale * x + shift), with a default of base == -1 to set the base to e, for the actual exp function.

[jeffdonahue]

Will probably merge this soon if nobody objects.

If you happened to look at this more than a few hours ago, I had added a caffe_*_add_scalar methods with different source/dest data pointers (the current version is "in-place"), but I've since removed that commit and changed to precomputing base^shift and scaling the exp result (since scal, unlike add_scalar, is actually a BLAS function so presumably faster).

[shelhamer]

Sounds good!

On Thu, Oct 2, 2014 at 5:52 PM, Jeff Donahue notifications@github.com
wrote:

Will probably merge this soon if nobody objects.

If you happened to look at this more than a few hours ago, I had added a
caffe_*_add_scalar methods with different source/dest data pointers (the
current version is "in-place"), but I've since removed that commit and
changed to precomputing base^shift and scaling the exp result (since scal,
unlike add_scalar, is actually a BLAS function so presumably faster).

—
Reply to this email directly or view it on GitHub
#1204 (comment).