Helio J. C. Barbosa

Artificial Immune Systems for Optimization

Artificial Immune Systems (AISs) are computational methods inspired by the biological immune system and thus classified as a nature-inspired meta-heuristic along with genetic algorithms, ant colony optimization, particle swarm optimization, and others. This chapter is focused on the application of AISs to solve optimization problems. Optimization is a mathematical principle largely applied to design and operational problems in all types of engineering, as well as a tool for formulating and solving inverse problems such as parameter identification in scientific and engineering situations. This chapter provides a survey of the applications of AISs in optimization. The main contributions are studied and contrasted with respect to the different concepts of the biological immune system, such as clonal selection, hypermutation, immune network, affinity, etc. that have inspired such techniques. The main types of optimization problems are considered, namely, (i) unconstrained problems, (ii) constrained problems, (iii) multimodal problems where several optima are to be obtained in a single run, and, finally (iv) multi-objective problems, where an approximation to the Pareto set is to be obtained as the result of a single run of the algorithm. Although AISs are good for solving optimization problems, useful features from other techniques have been combined with a “pure” AIS in order to generate hybrid AIS methods with improved performance.

A Critical Review of Adaptive Penalty Techniques in Evolutionary Computation

Constrained optimization problems are common in the sciences, engineering, and economics. Due to the growing complexity of the problems tackled, nature-inspired metaheuristics in general, and evolutionary algorithms in particular, are becoming increasingly popular. As move operators (recombination and mutation) are usually blind to the constraints, most metaheuristics must be equipped with a constraint handling technique. Although conceptually simple, penalty techniques usually require user-defined problem-dependent parameters, which often significantly impact the performance of a metaheuristic. A penalty technique is said to be adaptive when it automatically sets the values of all parameters involved using feedback from the search process without user intervention. This chapter presents a survey of the most relevant adaptive penalty techniques from the literature, identifies the main concepts used in the adaptation process, as well as observed shortcomings, and suggests further work in order to increase the understanding of such techniques.

Surrogate-Assisted Artificial Immune Systems for Expensive Optimization Problems

When an animal is exposed to antigens an efficient immune response is developed in order to defend the organism where specific antibodies are produced to combat them. The best antibodies multiply (cloning) and are improved (hypermutation and replacement) while new antibodies, produced by the bone marrow, are generated. Thus, if the organism is again attacked by the same antigen a quicker immune response takes place. This scheme of adaptation is known as clonal selection and affinity maturation by hypermutation or, more simply, clonal selection (Garrett, 2004). Computational methods inspired by the biological immune system are called Artificial Immune Systems (AISs). Immune-inspired algorithms have found applications in many domains. One of the most important area, the optimization, is a mathematical principle largely applied to design and operational problems in all types of engineering, as well as a tool for formulating and solving inverse problems such as parameter identification in scientific and engineering situations. When applied to optimization problems, the AISs are stochastic populational search methods which do not require a continuous, differentiable, or explicit objective function, and do not get easily trapped in local optima. However, the AISs, as well as other nature-inspired techniques, usually require a large number of objective function evaluations in order to reach a satisfactory solution. As modern problems have lead to the development of increasingly complex and computationally expensive simulation models, this becomes a serious drawback to their application in several areas such as Computational Structural Mechanics, Reservoir Simulation, Environmental Modeling, and Molecular Dynamics. Thus, a good compromise between the number of calls to the expensive simulation model and the quality of the final solutions must often be established. A solution to this problem is to modify the search process in order to obtain either a reduction on the total computational cost or an increase in the efficiency of the optimization procedure. The solution considered here is the use of a surrogate model (or metamodel), which provides an approximation of the objective function, replacing the computationally intensive original simulator evaluation. Additionally, the surrogate model can help to smooth out the objective function landscape, and facilitate the optimization process. The idea of reducing the computation time or improving the solutions performing less computationally expensive function evaluations can be found in the evolutionary computation literature (Bull, 1999; El-Beltagy et al., 1999; Jin, 2002; Zhou, 2004; Rasheed,

On GA-AIS Hybrids for Constrained Optimization Problems in Engineering

A genetic algorithm (GA) is hybridized with an artificial immune system (AIS) as an alternative to tackle constrained optimization problems in engineering. The AIS is inspired by the clonal selection principle and is embedded into a standard GA search engine in order to help move the population into the feasible region. The resulting GA-AIS hybrid is tested in a suite of constrained optimization problems with continuous variables, as well as structural and mixed integer reliability engineering optimization problems. In order to improve the diversity of the population, a variant of the algorithm is developed with the inclusion of a clearing procedure. The performance of the GA-AIS hybrids is compared with that of alternative techniques, such as the Adaptive Penalty Method, and the Stochastic Ranking technique, which represent two different types of constraint handling techniques that have been shown to provide good results in the literature.

Using Crowding-Distance in a Multiobjective Genetic Algorithm for Protein Structure Prediction

In this paper the insertion of the crowding-distance technique in a multiobjective genetic algorithm with phenotypic crowding is carried out for the protein structure prediction (PSP) problem. The main goal is obtain a more diversified and well distributed Pareto frontiers at the end of the optimization process. Three classical force field potentials, three hydrogen bond potentials and a hydrophobic compactation term were combined in two configurations with different objectives for the fitness function. A set of 45 proteins was used to evaluate the performance of the predictions. The results were compared against the previous mono- and multiobjective approaches, and with QUARK and MEAMT, two consolidated free-modeling PSP methodologies. The strategy proposed here was able to obtain improvements in the predicted models relative to the previous mono- and multiobjective approaches, proving to be quite promising in dealing with the PSP problem.

Predicting the Performance of Job Applicants by Means of Genetic Programming

Since their early development, genetic programming-based algorithms have been showing to be successful at challenging problems, attaining several human-competitive results and other awards. This paper will present another achievement of such algorithms by describing how our team has won an international machine-learning competition. We have solved, by means of grammar-based genetic programming techniques, a real-world problem of meritocracy in jobs by evolving classifiers that were both accurate and human-readable.

A parallel multi-GPU Clonal Selection Algorithm for optimization using OpenCL and OpenMP

Artificial Immune Systems are nature inspired techniques which have been applied with success to several areas. The Clonal Selection Algorithm (CLONALG) is one of the most used immune inspired techniques for optimization. Similarly to other metaheuristics, CLONALG requires a large number of objective function evaluations making it impracticable when the objective function is computationally expensive. One way to decrease the time required to solve a problem is by using parallel computing. In spite of CLONALG being easily parallelizable, few works can be found regarding this feature. A parallel CLONALG is proposed here using the OpenCL framework and the OpenMP library, making it possible to execute the search in parallel in different types of devices, such as CPUs, GPUs and APUs. Computational experiments are performed and the obtained results are compared with those found in the literature.

Strategies for Parallel Ant Colony Optimization on Graphics Processing Units

Ant colony algorithms are known to have a significant ability of finding high-quality solutions in a reasonable time [2]. However, the computational time of these methods is seriously compromised when the current instance of the problem has a high dimension and/or is hard to solve. In this line, a significant amount of research has been done in order to reduce computation time and improve the solution quality of ACO algorithms by using parallel computing. Due to the independence of the artificial ants, which are guided by an indirect communication via their environment (pheromone trail and heuristic information), ACO algorithms are naturally suitable for parallel implementation.

Multi- and Many-Threaded Heterogeneous Parallel Grammatical Evolution.

There are some algorithms suited for inference of human-interpretable models for classification and regression tasks in machine learning, but it is hard to compete with Grammatical Evolution (GE) when it comes to powerfulness, model expressiveness and ease of implementation. On the other hand, algorithms that iteratively optimize a set of programs of arbitrary complexity—which is the case of GE—may take an inconceivable amount of running time when tackling complex problems. Fortunately, GE may scale to such problems by carefully harnessing the parallel processing of modern heterogeneous systems, taking advantage of traditional multi-core processors and many-core accelerators to speed up the execution by orders of magnitude. This chapter covers the subject of parallel GE, focusing on heterogeneous multi- and many-threaded decomposition in order to achieve a fully parallel implementation, where both the breeding and evaluation are parallelized. In the studied benchmarks, the overall parallel implementation runtime was 68 times faster than the sequential version, with the program evaluation kernel alone hitting an acceleration of 350 times. Details on how to efficiently accomplish that are given in the context of two well-established open standards for parallel computing: OpenMP and OpenCL. Decomposition strategies, optimization techniques and parallel benchmarks followed by analyses are presented in the chapter.

Using grammar-based genetic programming to determine characteristics of multiple infections and environmental factors in the development of allergies and asthma

In recent decades asthma and allergies had great increase worldwide, being currently a serious global health problem. The causes of these disorders are unknown, but the most accepted hypothesis is that improving hygiene and reducing infections may be the main cause of this increase. Both asthma and allergies are complex diseases with strong environmental influence, so the use of versatile tools such as genetic programming can be important in the understanding of those conditions. We applied genetic programming to data obtained from 1296 children. Data related to chronic viral infections and environmental factors were used to classify in asthmatic and non-asthmatic, IgE and SPT in order to assess allergy. For asthma, viral infections were not relevant while for IgE and SPT they were. The use of genetic programming is shown to be a powerful tool to help understand those conditions.