General idea would be to provide a simple wrapper class for composing Models similar to the way a Model composes Layers. This would not provide new functionality, it would just be a way to make Layer grouping explicit for complex Models. I think this would be fairly straightforward to implement. This could also make separate fitting of Layer groups more intuitive.
Ex: split LN into two models
linear_nonlinear = MultiModel(
Model(WeightChannels(...), FIR(...)), # Linear part
Model(DoubleExponential(...)) # Nonlinear part
)
General idea would be to provide a simple wrapper class for composing Models similar to the way a Model composes Layers. This would not provide new functionality, it would just be a way to make Layer grouping explicit for complex Models. I think this would be fairly straightforward to implement. This could also make separate fitting of Layer groups more intuitive.
Ex: split LN into two models
Ex: model with multiple "branches"