Neural networking is the science of creating computational solutions modeled after the brain.  Like the human brain, neural networks are trainable-once they are taught to solve one complex problem, they can apply their skills to a new set of problems without having to start the learning process from scratch. 

The Lab seeks models which combine the best aspects of neural network mechanisms and symbolic artificial intelligence machine learning paradigms. Neural networks and machine learning algorithms represent a dramatic departure from conventional programming techniques. Rather than explicitly build a program to solve a problem, examples, called "training sets," of a type of problem are given, which the neural network "learns" how to solve. The network can then be presented with new examples on which it was not trained, known as "test sets," and it will use the skills it gained from the training set to formulate solutions. For many tasks, neural networks actually outperform human experts. For example, a doctor must go through years of training to learn to diagnose a disease on the basis of a set of symptoms. A neural network, in comparison, would diagnose a disease by first learning from a training set made up of symptoms with the correct diagnoses, and then would formulate a diagnosis when presented with new symptoms on which it was not trained.
Some of the problems considered in the Neural Networks and Machine Learning Laboratory are control problems, such as controlling a large flock of independent robots. Just as a person who understands how to drive a car can transfer that knowledge to driving a truck, a computer which controls one robot should quickly learn how to control a whole flock. Other problems being solved in the lab are planning and classification tasks. One project would allow computers to be able to recognize individuals' facial features and thus pick individuals out of photographs. Other applications include automatically sorting music libraries and classifying species of plants and animals.