Add more special feeders that take a specific selection of sources (with or without variable beta-like weight factors applied), use it as input for an arbitrary N->1 dimensional function (e.g. sum, dot_product, euclidean distance, ..) and finally optionally apply an activation function.
Add more special feeders that take a specific selection of sources (with or without variable beta-like weight factors applied), use it as input for an arbitrary N->1 dimensional function (e.g. sum, dot_product, euclidean distance, ..) and finally optionally apply an activation function.