On the fly expansion of linear layers

A linear layer takes a set of inputs, and for each node in its layer it sums weights*inputs to calculate the node output. This output is then fed to the next layer (or is the network output if it's the final layer). Neural network architectures are ...