I want to use an Elementwise max layer in a caffe network. Looking at the available layers I could not find a suitable substitute. I wonder if someone has already done something in this direction, or has an idea how to go about implementing this type of layer in caffe.

My current plan is to use a concatenate layer followed by within-channel max pooling, even though, adding the within channel pooling feature seems a bit daunting to me. Is this a reasonable approach?

EDIT: I just realize I missed looking at the elementwise layer. Adding a max case is all I really need to do.