Connectionist Systems for Fishing Prediction

This chapter explains the foundations of a new support system for fisheries, based on connectionist techniques, digital image treatment, and fuzzy logic. The purpose of our system is to increase the output of the pelagic fisheries without endangering the natural balance of the fishing resources. It uses data from various remote sensors and the logbook of a collaborating fishing boat to improve the catches of the Prionace Glauca, a pelagic shark species also known as the blue shark.

Neuroglial Behaviour in Computer Science

This chapter presents a study that incorporates into the connectionist systems new elements that emulate cells of the glial system. More concretely, we have considered a determined category of glial cells known as astrocytes, which are believed to be directly implicated in the brain’s information processing. Computational models have helped to provide a better understanding of the causes and factors that are involved in the specific functioning of particular brain circuits. The present work will use these new insights to progress in the field of computing sciences and artificial intelligence. The proposed connectionist systems are called artificial neuroglial networks (ANGN).

Cooperative AI Techniques for Stellar Spectra Classification

This chapter describes a hybrid approach to the unattended classification of low-resolution optical spectra of stars. The classification of stars in the standard MK system constitutes an important problem in the astrophysics area, since it helps to carry out proper stellar evolution studies. Manual methods, based on the visual study of stellar spectra, have been frequently and successfully used by researchers for many years, but they are no longer viable because of the spectacular advances of the objects collection technologies, which gather a huge amount of spectral data in a relatively short time. Therefore, we propose a cooperative system that is capable of classifying stars automatically and efficiently, by applying to each spectrum the most appropriate method or combined methods, which guarantees a reliable, consistent, and adapted classification. Our final objective is the integration of several artificial intelligence techniques in a unique hybrid system.

Predicting Credit Ratings with a GA-MLP Hybrid

The practical application of MLPs can be time-consuming due to the requirement for substantial modeler intervention in order to select appropriate inputs and parameters for the MLP. This chapter provides an example of how elements of the task of constructing a MLP can be automated by means of an evolutionary algorithm. A MLP whose inputs and structure are automatically selected using a genetic algorithm (GA) is developed for the purpose of predicting corporate bond-issuer ratings. The results suggest that the developed model can accurately predict the credit ratings assigned to bond issuers.

Time Series Forecasting by Evolutionary Neural Networks

This chapter presents a hybrid evolutionary computation/neural network combination for time series prediction. Neural networks are innate candidates for the forecasting domain due to advantages such as nonlinear learning and noise tolerance. However, the search for the ideal network structure is a complex and crucial task. Under this context, evolutionary computation, guided by the Bayesian Information Criterion, makes a promising global search approach for feature and model selection. A set of 10 time series, from different domains, were used to evaluate this strategy, comparing it with a heuristic model selection, as well as with conventional forecasting methods (e.g., Holt-Winters & Box-Jenkins methodology).

Soft Computing Approach for Bond Rating Prediction

Soft computing is popularly referred to as a collection of methodologies that work synergistically and provide flexible information processing capabilities for handling real-life situations. Its aim is to exploit the tolerance for imprecision, uncertainty and approximate reasoning in order to achieve tractability and robustness. Currently, fuzzy logic, artificial neural networks, and genetic algorithms are three main components of soft computing. In this chapter, we show the application of soft computing techniques to solve high dimensional problems. We have taken a multi-class classification problem of bond rating prediction with 45 input variables and have used soft computing techniques to solve it. Two techniques, namely dimensionality reduction technique and variable reduction techniques, have been tried and their performances are compared. Hybrid networks are found to give better results compared to normal fuzzy and ANN methods. We also have compared all the results with normal regression techniques.

Several Approaches to Variable Selection by Means of Genetic Algorithms

This chapter shows several approaches to determine how the most relevant subset of variables can perform a classification task. It will permit the improvement and efficiency of the classification model. A particular technique of evolutionary computation, the genetic algorithms, is applied which aim to obtain a general method of variable selection where only the fitness function will be dependent on the particular problem. The solution proposed is applied and tested on a practical case in the field of analytical chemistry to classify apple beverages.

Using Genetic Programming to Extract Knowledge from Artificial Neural Networks

A technique is described in this chapter that makes it possible to extract the knowledge held by previously trained artificial neural networks. This makes it possible for them to be used in a number of areas (such as medicine) where it is necessary to know how they work, as well as having a network that functions. This chapter explains how to carry out this process to extract knowledge, defined as rules. Special emphasis is placed on extracting knowledge from recurrent neural networks, in particular when applied in predicting time series.

Self-Adapting Intelligent Neural Systems Using Evolutionary Techniques

This chapter describes genetic algorithm-based evolutionary techniques for automatically constructing intelligent neural systems. These techniques can be used to build and train multilayer perceptrons with the simplest architecture. These neural networks are usually designed using binary-coded genetic algorithms. The authors show how the basic architectures codification method, which uses an algebra-based codification, employs a shorter string length and voids illegal architectures in the search space. The networks are trained using real number codification. The morphological crossover operator is presented and compared to other important real-coded crossover operators. The purpose is to understand that the combination of all these techniques results in an evolutionary system, which self-adaptively constructs intelligent neural systems to solve a problem given as a set of training patterns. To do so, the evolutionary system is applied in laboratory tests and to a real-world problem: breast cancer diagnosis.

Development of ANN with Adaptive Connections by CE

This chapter presents the use of Artificial Neural Networks (ANN) and Evolutionary Computation (EC) techniques to solve real-world problems including those with a temporal component. The development of the ANN maintains some problems from the beginning of the ANN field that can be palliated applying EC to the development of ANN. In this chapter, we propose a multilevel system, based on each level in EC, to adjust the architecture and to train ANNs. Finally, the proposed system offers the possibility of adding new characteristics to the processing elements (PE) of the ANN without modifying the development process. This characteristic makes possible a faster convergence between natural and artificial neural networks.