General MLP NN has different structure on each layer

In[]:=
[◼]
NetGraphPlot
[NetTrain[NetChain[{6,ElementwiseLayer["SELU"],6,ElementwiseLayer["SELU"],6,ElementwiseLayer["SELU"],{}},"Input"->"Real"],Thread[#->Sin[#]]&@Range[0,10,0.1]],ImageSize->180]
Out[]=

Can we make a CA-like NN?

ConvolutionLayer[]

Want a neural net layer where the data is repeatedly passed through the layer...

ICCA : inhomogeneous continuous CA

E.g. ramp + antiramp [i.e. change activation functions at each cell]