Ensembles of Neural Networks for Digital Problems

Abstract

Ensembles of neural networks is a technique that uses several different networks working together to find a relationship that exactly matches a many-to-one training set for digital classification problems. Each network learns a different region of the training space and all these regions fit together, like pieces of a jigsaw puzzle, to cover the entire training space. The individual networks are ‘grown’ as they are needed to form either cascades or branches of networks. The networks can be of any type such as backpropagation, cascade etc. However, virtually any other technique can be used in place of the networks: GA, EA, DRS, tabu search, nearest neighbour, and so on. Methods are discussed to improve the generalisation.