In a Neural network, there is an input layer, any number of hidden layers, and an output layer. My question is: Are the input and output layer nodes actually perceptions? Or do they just signify what/how many/where the inputs and outputs are?
Not 100% sure I got your question right, but yes the input nodes are perception fields. For the output it's good if you choose a space such that the different outputs you expect are independent.
One method to improve neural networks is actually using receptive fields. Look for convolutionary networks and pooling.
Would have posted that to a comment if I could.