Learning and Generalization Controlled by Contradiction

SG (Specific to General) is a network that learns from a training set containing examples. Each example gives an input pattern along with the output that the network should produce for that input. The training set is a subset of the complete mapping from input to output. Therefore, the network should not only converge to a representation that contains the information given by the training set, but also generalize that information so that the network will respond well to inputs that it has not been trained on.