Segmented Neural Network Modelling Using Adaboost

The objective of this paper was to model a non-linear system, which could ordinarily be modelled using a large complicated neural network, using a combination of smaller more simple neural networks without loosing too much information.This objective was achieved through the research and use of several mathematical techniques including boosting and back-propagation to develop an algorithm which could extract the linear features of the non-linear system. Implementation of this algorithm was then done using Matlab as a modelling tool, this includes useful tools for dealing with neural networks.The results shown suggest that this technique can indeed model a complicated system by "breaking" it into smaller simpler sections.