How I can implement a custom hidden and a regression output layer with 2 inputs?

I want bulid up with the deep learning toolbox (CNN, deppNrworkDesigner, layerGraph) a variational autoencoder for a sequence data input.

For this I need one custom layer with 2 inputs and a custom regression output layer in my networtk. But the documentation page "Define Custome Deep Learning Layers" just explain templates for layers with one input.

Howe i can adapt this templates for layers with more than one input and how i can use this layer in functions and apps (deepNetworkDesigner, trainNetwork, layerGraph-Functions etc.)?

i have Adapt the class of the Addition Layer, but it doesnt work well.

Answer 1

[Greg Heath]

Is that one input a scalar or a vector?

Hope this helps

**Thank you for formally accepting my answer**

Greg

[Markus Wagner]

The input of the network is a vector. The inputs of the layer are vectors.