Juan Rios

Evolutionary system for automatically constructing and adapting radial basis function networks

Abstract This article presents a new system for automatically constructing and training radial basis function networks based on original evolutionary computing methods. This system, called Genetic Algorithm Radial Basis Function Networks (GARBFN), is based on two cooperating genetic algorithms. The first algorithm uses a new binary coding, called basic architecture coding, to get the neural architecture that best solves the problem. The second, which uses real coding, takes its inspiration from mathematical morphology theory and trains the architectures output by the binary genetic algorithm. This system has been applied to a laboratory problem and to breast cancer diagnosis. The results of these evaluations show that the overall performance of GARBFN is better than other related approaches, whether or not they are based on evolutionary techniques.

Crossover and mutation operators for grammar-guided genetic programming

This paper proposes a new grammar-guided genetic programming (GGGP) system by introducing two original genetic operators: crossover and mutation, which most influence the evolution process. The first, the so-called grammar-based crossover operator, strikes a good balance between search space exploration and exploitation capabilities and, therefore, enhances GGGP system performance. And the second is a grammar-based mutation operator, based on the crossover, which has been designed to generate individuals that match the syntactical constraints of the context-free grammar that defines the programs to be handled. The use of these operators together in the same GGGP system assures a higher convergence speed and less likelihood of getting trapped in local optima than other related approaches. These features are shown throughout the comparison of the results achieved by the proposed system with other important crossover and mutation methods in two experiments: a laboratory problem and the real-world task of breast cancer prognosis.

Initialization method for grammar-guided genetic programming

This paper proposes a new tree-generation algorithm for grammar-guided genetic programming that includes a parameter to control the maximum size of the trees to be generated. An important feature of this algorithm is that the initial populations generated are adequately distributed in terms of tree size and distribution within the search space. Consequently, genetic programming systems starting from the initial populations generated by the proposed method have a higher convergence speed. Two different problems have been chosen to carry out the experiments: a laboratory test involving searching for arithmetical equalities and the real-world task of breast cancer prognosis. In both problems, comparisons have been made to another five important initialization methods.

Optimisation With Real-Coded Genetic Algorithms Based On Mathematical Morphology

The goal of this work is to propose a novel approach to function optimisation by evolutionary techniques, in particular, real-coded genetic algorithms. A new genetic crossover operator, suitable for real codification, has been designed. This operator is called morphological crossover as it is based on mathematical morphology theory. The morphological crossover includes a new genetic diversity measure that has low computational cost. This operator is presented along with the resolution of a set of optimisation problems, including neural network training. The results are compared to other optimisation approaches as gradient descent methods or binary and real-coded genetic algorithms using different crossover operators. These tests show that the properties exhibited by the proposed operator when using real-coded genetic algorithms give higher convergence speed and less probability of being trapped in a local optimum.

Evolutionary construction and adaptation of intelligent systems

This paper introduces evolutionary techniques for automatically constructing intelligent self-adapting systems, capable of modifying their inner structure in order to learn from experience and self-adapt to a changing environment. These evolutionary techniques comprise an evolutionary system that is engineered by grammar-guided genetic programming, enabling the development of sub-symbolic and symbolic intelligent systems: artificial neural networks and knowledge-based systems, respectively. A context-free-grammar based codification system for artificial neural networks and rules, an initialization method and a crossover operator have been designed to properly balance the exploration and exploitation capabilities of the proposed system. This speeds up the convergence process and avoids trapping in local optima. This system has been applied to a medical domain: the detection of knee injuries from the analysis of isokinetic time series. The results of the evolved symbolic and sub-symbolic intelligent systems have been statistically compared with each other as part of a quantitative and qualitative performance analysis.

Cooperative binary-real coded genetic algorithms for generating and adapting artificial neural networks

This paper describes a genetic system for designing and training feed-forward artificial neural networks to solve any problem presented as a set of training patterns. This system, called GANN, employs two interconnected genetic algorithms that work parallelly to design and train the better neural network that solves the problem. Designing neural architectures is performed by a genetic algorithm that uses a new indirect binary codification of the neural connections based on an algebraic structure defined in the set of all possible architectures that could solve the problem. A crossover operation, known as Hamming crossover, has been designed to obtain better performance when working with this type of codification. Training neural networks is also accomplished by genetic algorithms but, this time, real number codification is employed. To do so, morphological crossover operation has been developed inspired on the mathematical morphology theory. Experimental results are reported from the application of GANN to the breast cancer diagnosis within a complete computer-aided diagnosis system.

Grammar based crossover operator in genetic programming

This paper introduces a new crossover operator for the genetic programming (GP)paradigm, the grammar-based crossover (GBX). This operator works with any grammar-guided genetic programming system. GBX has three important features: it prevents the growth of tree-based GP individuals (a phenomenon known as code bloat), it provides a satisfactory trade-off between the search space exploration and the exploitation capabilities by preserving the context in which subtrees appear in the parent trees and, finally, it takes advantage of the main feature of ambiguous grammars, namely, that there is more than one derivation tree for some sentences (solutions). These features give GBX a high convergence speed and low probability of getting trapped in local optima, as shown throughout the comparison of the results achieved by GBX with other relevant crossover operators in two experiments: a laboratory problem and a real-world task: breast cancer prognosis.

Iterative method for automatic detection of masses in digital mammograms for computer-aided diagnosis

An iterative algorithm has been developed for automatic detection of breast masses from digitalized mammograms. The procedure has been divided in two stages. The first one based on the histogram analysis of the input image. The second one employs a topological analysis from the results obtained in the first stage. The final output is a set of interest regions that are defined as suspicious areas by the system. These suspicious regions should be harder studied in order to present a final diagnosis. The developed system may be used together with any other suspicious area diagnosis algorithms. In this way a computer assisted diagnosis (CAD) program to assist radiologists in his mammography interpretation task could be easy developed.

Real-Coded Genetic Algorithms Based on Mathematical Morphology

The goal of this work is to propose a general-purpose crossover operator for real-coded genetic algorithms that is able to avoid the major problems found in this kind of approach such as the premature convergence to local optima, the weakness of genetic algorithms in local fine-tuning and the use of realcoded genetic algorithms instead of the traditional binary-coded problems. Mathematical morphology operations have been employed with this purpose adapting its meaning from other application fields to the generation of better individuals along the evolution in the convergence process. This new crossover technique has been called mathematical morphology crossover (MMX) and it is described along with the resolution of systematic experiments that allow to test its high speed of convergence to the optimal value in the search space.

Evolutionary local search of fuzzy rules through a novel neuro-fuzzy encoding method

This paper proposes a new approach for constructing fuzzy knowledge bases using evolutionary methods. We have designed a genetic algorithm that automatically builds neuro-fuzzy architectures based on a new indirect encoding method. The neuro-fuzzy architecture represents the fuzzy knowledge base that solves a given problem; the search for this architecture takes advantage of a local search procedure that improves the chromosomes at each generation. Experiments conducted both on artificially generated and real world problems confirm the effectiveness of the proposed approach.