Bandyopadhyay S., Pal S.K. Classification and Learning Using Genetic Algorithms. Applications in Bioinformatics and Web Intelligence

Springer, 2007. — 320 p.Genetic algorithms (GAs) are randomized search and optimization techniques guided by the principles of evolution and natural genetics; they have a large amount of implicit parallelism. GAs perform multimodal search in complex landscapes and provide near-optimal solutions for objective or fitness function of an optimization problem. They have applications in fields as diverse as pattern recognition, image processing, neural networks, machine learning, jobshop scheduling and VLSI design, to mention just a few.
Traditionally, GAs were designed to solve problems with an objective to optimize only a single criterion. However, many real-life problems involve mul- tiple conicting measures of performance, or objectives, which need simultaneous optimization. Optimal performance according to one objective, if such an optimum exists, often implies unacceptably low performance in one or more of the other objective dimensions, creating the need for a compromise to be reached. A suitable set of solutions, called the Pareto-optimal set, to such problems is one where none of the solutions can be further improved on any one objective without degrading it in another. In recent times, there has been a spurt of activities in exploiting the significantly powerful search capability of GAs for multiobjective optimization, leading to the development of several algorithms belonging to the class of multiobjective genetic algorithms (MOGAs).
The present volume is aimed at providing a treatise in a unified framework, with both theoretical and experimental results, describing the basic principles of GAs and MOGAs, and demonstrating the various ways in which genetic learning can be used for designing pattern recognition systems in both supervised and unsupervised modes. Their applications to bioinformatics and Web mining are also described. The task of classification, an integral part of pattern recognition, can be viewed as a problem of generating appropriate class boundaries that can successfully distinguish the various classes in the feature space. In real-life problems, the boundaries between different classes are usually complex and nonlinear. It is known that any nonlinear surface can be approximated by a number of simpler lower-order surfaces. Hence the problem of classification can be viewed as searching for a number of simpler surfaces (e.g., hyperplanes) that can appropriately model the class boundaries while providing the minimum number of misclassified data points. It is shown how GAs can be employed for approximating the class boundaries of a data set such that the recognition rate of the resulting classifier is sometimes com- parable to, and often better than, several widely used classification methods including neural networks, for a wide variety of data sets. Theoretical analysis of the classifier is provided. The effectiveness of incorporating the concept of chromosome differentiation in GAs, keeping analogy with the sexual differentiation commonly observed in nature, in designing the classifier is studied. Since the classification problem can be naturally modelled as one of multiobjective optimization, the effectiveness of MOGAs is also demonstrated in this regard.
Clustering is another important and widely used exploratory data analysis tool, where the objective is to partition the data into groups such that some similarity/dissimilarity metric is optimized. Since the problem can be posed as one of optimization, the application of GAs to it has attracted the attention of researchers. Some such approaches of the application of GAs to clustering, both crisp and fuzzy, are described.Genetic Algorithms
Supervised Classification Using Genetic Algorithms
Theoretical Analysis of the GA-classifier
Variable String Lengths in GA-classifier
Chromosome Differentiation in VGA-classifier
Multiobjective VGA-classifier and Quantitative Indices
Genetic Algorithms in Clustering
Genetic Learning in Bioinformatics
Genetic Algorithms and Web Intelligence
A: ϵ-Optimal Stopping Time for GAs
B: Data Sets Used for the Experiments
C: Variation of Error Probability with P₁