Neural Networks - Model Selection

In order to select the appropriate neural network model, it is necessary to find out whether you are dealing with a
 

• supervised task,
• or unsupervised task.

If you want to train the system to respond with a certain predefined output, the task is supervised. If the network should self-organize, the target is not predetermined and the task is called an unsupervised task.

Furthermore, you have to know whether your task is a
 

•  classification task,
• or function approximation task.

A neural network performing a classification task puts each input into one of a given number of classes. For function approximation tasks, each input is associated with a certain (analog) value.