How do we determine the number of hidden layers? In the slides ahead, we always assume that we've created a network with N number of hidden layers and then simply tune the parameters.

Do we have iterative algorithms to determine the number of necessary hidden layers for a given problem or does the programmer/researcher dictate it?

I did a little bit of poking around on stack-overflow and Quora, and it seems that determining the optimal architecture for a neural network beforehand is a very difficult problem theoretically, although in practice it seldom comes up. The layout of your input and output layers are determined by the problem, so it's just a matter of determining the number of hidden layers and the number of nodes in those layers. For this part, it seems that putting in excess layers and excess nodes per layer and using a pruning algorithm during training results in more optimal results than trying to noodle it out beforehand and statically specifying the architecture.

http://stackoverflow.com/questions/3345079/estimating-the-number-of-neurons-and-number-of-layers-of-an-artificial-neural-ne