I've been reading up on artificial neural networks, and a couple things are not quite clear to me. I'm hoping to make my own simple feedforward neural network just for practice.
So first, do neurons always pass on their values to the neurons in the next layers, or do they only fire when they are "stimulated" enough and have a certain connection weight or something?
Second (most importantly), are the weights stored with the neurons or with the connections between the neurons?
And third - with backpropagation, how do I go back and tell exactly what neuron or connection (whichever has the weight) I need to adjust?