| The current generation of non-modular neural network classifiers is unable to cope with classification problems which have a wide range of overlap among classes. This is due to the high coupling among the network's hidden nodes. Modular neural network structures attempt to reduce this limitation via a divide-and-conquer approach. However, current modular designs are not offering a general alternative to the non-modular approach because they do not provide a reasonable balance between sub-tasks simplification, and decision-making efficiency. When the task-decomposition algorithm attempts to produce sub-tasks as simple as they can be, the modules are expected to give the multi-module decision-making strategy enough information to take an accurate global decision.;This thesis proposes a new modular neural network design for classification. The new model outperforms the state-of-the-art modular and non-modular neural classifiers using the same supervised learning scheme. Its success is due to its multi-stage task-decomposition according to the different levels of overlap among classes, cooperation among its modules by "voting," the ability of its "specialized-modules" to resolve high overlaps, and finally, the more "balance" it provides between the simplification of sub-tasks and the efficiency of the multi-module decision-making strategy. The performance of the proposed model is assessed by testing it on several benchmark applications. The results are compared to those of several alternative modular and non-modular neural classifiers. |
