Database Analysis with ANNs by means of Graph Evolution

Database Analysis with ANNs by means of Graph Evolution

Daniel Rivero (University of A Coruña, Spain), Julián Dorado (University of A Coruña, Spain), Juan Rabuñal (University of A Coruña, Spain) and Alejandro Pazos (University of A Coruña, Spain)
Copyright: © 2011 |Pages: 15
DOI: 10.4018/978-1-60960-102-7.ch005

Abstract

Traditionally, the development of Artificial Neural Networks (ANNs) is a slow process guided by the expert knowledge. This expert usually has to test several architectures until he finds one suitable for solving a specific problem. This makes the development of ANNs a slow process in which the expert has to do much effort. This chapter describes a new method for the development of Artificial Neural Networks, so it becomes completely automated. Since ANNs are complex structures with very high connectivity, traditional algorithms are not suitable to represent them. For this reason, in this work graphs with high connectivity that represent ANNs are evolved. In order to measure the performance of the system and to compare the results with other ANN development methods by means of Evolutionary Computation (EC) techniques, several tests were performed with problems based on some of the most used test databases in Data Mining. These comparisons show that the system achieves good results that are not only comparable to those of the already existing techniques but, in most cases, improve them.
Chapter Preview
Top

Background

Genetic Programming

GP (Koza, 92) is based on the evolution of a given population. Its working is similar to a GA. In this population, every individual represents a solution for a problem that is intended to be solved. The evolution is achieved by means of the selection of the best individuals – although the worst ones have also a little chance of being selected – and their mutual combination for creating new solutions. This process is developed using selection, crossover and mutation operators. After several generations, the population is expected to contain some good solutions for the problem.

The GP encoding for the solutions is tree-shaped, so the user must specify which are the terminals (leaves of the tree) and the functions (nodes that have children) for being used by the evolutionary algorithm in order to build complex expressions. These can be mathematical (including, for instance, arithmetical or trigonometric operators), logical (with Boolean or relational operators, among others) or other type of even more complex expressions.

The wide application of GP to various environments and its consequent success are due to its capability for being adapted to numerous different problems. Although the main and more direct application is the generation of mathematical expressions (Rivero, 2005), GP has been also used in others fields such as filter design (Rabuñal, 2003), knowledge extraction (Rabuñal, 2004), image processing (Rivero, 2004), etc.

Purchase this chapter to continue reading all 15 pages >

Complete Chapter List

Search this Book:
Reset