Neural Network Training

[Niven]

I'm trying to learn about neural networks and AIs. So far, I think I'm understanding most everything but I am having trouble figuring out one thing. How do I teach a neural network(feedforward network)? For example, let's say I have a three layer network with 2 perceptrons in the input layer, 4 in the hidden layer, and 1 in the output layer, and I know the desired output of the output layer only, how do I adjust the weights of all of the other layers? Do I need to know the desired output for each individual neuron (which I hope I don't because that would get annoying quick with larger nets) or is it possible to do it off of only knowing the final desired output(s)?

[Peter87]

You only need to know the desired output of the output layer. A common method to train feedforward neural networks is back-propagation algorithm.
http://en.wikipedia.org/wiki/Backpropagation

[DeXecipher]

All i know so far is that it is described using graph theory and matrix transformations, finding the shortest path

http://en.wikibooks.org/wiki/Artificial_Intelligence/Search/Dijkstra%27s_Algorithm

[Niven]

Thanks for the replies. I'm going to check out those methods and see what I can do with them.