Evolving Objects

J. J. Merelo, J. Carpio, P. Castillo, V. M. Rivas, G. Romero
GeNeura Team
todos@geneura.ugr.es, http://geneura.ugr.es

Abstract:

This paper presents evolving objects (EO), a new framework for evolutionary computation (EC) that tries to unify all current paradigms, and free them from some limitations, allowing to integrate evolutionary computation into current operating systems and networks. Unlike current EC paradigms, such as genetic algorithms or evolution strategies, which concentrate on representation, EO concentrates on interfaces; any object can evolve if it is endowed with an interface to do so. In this paper, we try to put EC into a wider perspective, and then cast it to an object oriented computation framework; we describe what features must have an object in order to evolve, and some examples of how EO has been put to practice evolving neural networks and Mastermind solutions, using a C++ class library that implements EO's ideas.