Neural Networks are a handy tool for every robot builder to know. They provide a relatively simple way of making a robot learn from past mistakes and successes. IBM developerWorks has added a nice tutorial on Neural Networks. The article provides background information on how backpropagation nets work and why the perceptron model doesn't work. A sample network is developed that can be trained to distinguish between four different programming languages by reading samples of the code.