I'm curious if the self-attention layer described in this paper could be implemented in MATLAB. Would it require a custom layer to construct? Does not seem feasible inside the Deep Network Designer. I'd appreciate a pointer in the right direction.
You would have to create a custom layer. If you are using R2020a version, you'll be able to load the custom layer into the Deep Network Designer from the workspace.
You can refer to the following link for information on creating a custom layer:
You can also select a web site from the following list:
How to Get Best Site Performance
Select the China site (in Chinese or English) for best site performance. Other MathWorks country sites are not optimized for visits from your location.
0 Comments
Sign in to comment.