Tree Architecture Pattern Distributor: A Task Decomposition Classification Approach

Abstract — Task decomposition is a widely used method to solve complex and large problems. In this paper, it is proposed a novel task decomposition approach, named Tree Architecture Pattern Distributor (TreeArchPD), which is based on another task decomposition technique, called Pattern Distributor. The main idea is to design a tree architecture with many Distributors instead of using only one Distributor as proposed by the original technique. It is also proposed a new class grouping method that aims to optimize the class selection for task decomposition. Many experiments were done and they showed the effectiveness of the proposed approaches. MULTILAYERED feed-forward neural networks arewidely used in the literature to solve diverse prob-lems. It is widely known that these networks present some drawbacks when applied in real world problems. A common drawback is the network size. Small networks cannot lear