Cursos de Verano de El Escorial:
Encuentro sobre Computación Natural


The aim of this presentation is to justify why Evolutionary Computation is a good compromise solution for the design and management of current telecommunication systems.

These systems have reached a level of complication and complexity such that classic analytic approaches to their building and maintenance do not provide feasible solutions anymore.

Among the random-component based algorithms that do not preclude a good knowledge of the space of possible solutions we find the population-based evolutionary algorithms. These are inspired by the concept of "survival of the fittest" in which fitter solutions to a specific problem get a better chance of reproducing, thus surviving, into future generations until one solution, hopefully, is good enough for our problem at hand.

In this presentation we won´t talk about representation models, variation and selection strategies, fitness landscapes and not even about different theoretical classes of evolutionary algorithms. We will concentrate on showing the audience how these algorithms, in general, can de used to design and manage telecom systems that display properties of adaptability, robustness and evolvability.

These three qualities are abundantly found in biological systems. The point is that current telecom systems suffer from a series of problems (in hardware and in software) that could be conquered if these systems were robust, adaptable and evolvable. We will give a brief description of these problems.

To start displaying some of these properties, we will start with a very simple example, the "ants algorithm". This system is adaptable and robust (but not evolvable) and is described by a local random-based algorithm, not an evolutionary algorithm as such.

We will continue with an artificial life system that allows for evolvability too. In a scale of goal-directed to not having a specific goal but reproduction and from single fixed solution to ecosystem-population based solution, the ants algorithm and alife systems are two extreme examples.

It is in the middle ground where we find the standard evolutionary algorithms and where most of the current applications to Telecom are derived from. In this presentation we will analyze the optimization in the design of a network with a genetic algorithm and the message filtering using genetic programming.

Página principal del curso