Catálogo de publicaciones - libros

This paper describes a new technique for automatically developing Artificial Neural Networks (ANNs) by means of an Evolutionary Computation (EC) tool, called Genetic Programming (GP). This paper also describes a practical application in the field of Data Mining. This application is the Iris flower classification problem. This problem has already been extensively studied with other techniques, and therefore this allows the comparison with other tools. Results show how this technique improves the results obtained with other techniques. Moreover, the obtained networks are simpler than the existing ones, with a lower number of hidden neurons and connections, and the additional advantage that there has been a discrimination of the input variables. As it is explained in the text, this variable discrimination gives new knowledge to the problem, since now it is possible to know which variables are important to achieve good results.

As an alternative to traditional artificial neural network approaches to pattern recognition, a hardware-implemented evolvable characters recognizer is presented in this paper. The main feature of the proposed evolvable system is that all the components including the evolutionary algorithm (EA), fitness calculation, and virtual reconfigurable circuit are implemented in a Xilinx Virtex xcv2000E FPGA. This allows for a completely pipelined hardware implementation and yields a significant speedup in the system evolution. In order to optimize the performance of the evolutionary algorithm and release the users from the time-consuming process of mutation parameters tuning, a self-adaptive mutation rate control scheme is also introduced. An analysis of experimental results demonstrates that the proposed evolvable system using self-adaptive mutation rates is superior to traditional fixed mutation rate-based approaches.

The prediction of operons is critical to reconstruction of regulatory networks at the whole genome level. In this paper, a multi-approach guided genetic algorithm is developed to prediction of operon. The fitness function is created by using intergenic distance of local entropy-minimization, participation of the same metabolic pathway, log-likelihood of COG gene functions and correlation coefficient of microarray expression data, which have been used individually for predicting operons. The gene pairs within operons have high fitness value by using these four scoring criteria, whereas those across transcription unit borders have low fitness value. On the other hand, it is easy to predict operons and makes the prediction ability stronger by using these four scoring criteria. The proposed method is examined on 683 known operons of and an accuracy of 85.9987% is obtained.

In this paper, an optimal direct design method of steel frame using advanced analysis and genetic algorithm is presented. The advanced analysis realistically assesses both strength and behavior of a structural system and its component members in a direct manner. The micro-GA is used for minimum weight optimization of steel frames. Constraint functions are load-carrying capacities and serviceability. The optimum designs determined by the proposed method are lighter than those given by other approaches.

Under the multi-region and multi-sector consideration, the previous double-objective optimal public investment model is extended to involve optimal employment rate objective and time-flow total income maximization objective first. Then genetic algorithm is applied to solve the multi-objective model. Finally a case study is carried out to verify the superiority of the genetic algorithm-based solution to traditional public investment distribution approach.

Many-objective optimization refers to multi-objective optimization problems with a number of objectives considerably larger than two or three. This papers contributes to the use of Particle Swarm Optimization (PSO) for the handling of such many-objective optimization problems. Multi-objective PSO approaches typically rely on the employment of a so-called set of leaders that generalizes the global best particle used in the standard PSO algorithm. The exponentially decreasing probability of finding non-dominated points in search spaces with increasing number of objectives poses a problem for the selection from this set of leaders, and renders multi-objective PSOs easily unusable. Gradual Pareto dominance relation can be used to overcome this problem. The approach will be studied by means of the problem to minimize the Euclidian distances to a number of points, where each distance to the points is considered an independent objective. The Pareto set of this problem is the convex closure of the set of points. The conducted experiments demonstrate the usefulness of the proposed approach and also show the higher resemblance of the proposed PSO variation with the standard PSO.

In this paper, a mixed integer programming method based on ant colony optimization is presented, and applied to the classical Unit Commitment problem. The idea is to reformulate the problem into a graph exploration structure, and to use discrete ant colony optimization to explicitly take into account time down, time up and demand constraints in the optimization procedure. This method is coupled with a continuous ant colony algorithm to compute produced powers. Results, obtained on relatively small cases, show the viability of the proposed approach: a near optimal solution, with guarantees of feasibility, can be computed with low computation times.

A shuffled complex evolution of particle swarm optimization algorithm called SCE-PSO is introduced in this paper. In the SCE-PSO, a population of points is sampled randomly in the feasible space. Then the population is partitioned into several complexes, which is made to evolve based on PSO. At periodic stages in the evolution, the entire population is shuffled and points are reassigned to complexes to ensure information sharing. Both theoretical and numerical studies of the SCE-PCO are presented. Five optimization problems with commonly used functions are utilized for evaluating the performance of the proposed algorithm, and the performance of the proposed algorithm is compared to PSO to demonstrate its efficiency.

This paper proposes a wasp swarm optimization algorithm, which is applied to the dynamic variant of the maximum satisfiability problem, or MAX-SAT. Here, we describe the changes implemented to optimize the dynamic problem and analyze the parameters of the new algorithm. Wasp swarm optimization accomplishes very well the task of adapting to systematic changes of dynamic MAX-SAT instances derived from static problems, and significantly outperforms the local search algorithm used as benchmark.

The multidimensional 0/1 knapsack problem is a classical problem of discrete optimization. There are several approaches for solving the different variations of such problem, including mathematical programming and stochastic heuristic methods. This paper presents the application of Particle Swarm Optimization (PSO) for the problem, using selected instances of ORLib. For the instances tested, results were very close or equal to the optimal solution known, even considering that the parameters of PSO were not optimized. The analysis of the results suggests the potential of a simple PSO for this class of combinatorial problems.