With regards to distributed processing of neural networks, how do you split up the neural network onto the various nodes? Do you cut it up into sections and then compute the combined weights from each nodes outputs?