Blended biogeography-based optimization for constrained optimization

Biogeography-based optimization (BBO) is a new evolutionary optimization method that is based on the science of biogeography. We propose two extensions to BBO. First, we propose a blended migration operator. Benchmark results show that blended BBO outperforms standard BBO. Second, we employ blended BBO to solve constrained optimization problems. Constraints are handled by modifying the BBO immigration and emigration procedures. The approach that we use does not require any additional tuning parameters beyond those that are required for unconstrained problems. The constrained blended BBO algorithm is compared with solutions based on a stud genetic algorithm (SGA) and standard particle swarm optimization 2007 (SPSO 07). The numerical results demonstrate that constrained blended BBO outperforms SGA and performs similarly to SPSO 07 for constrained single-objective optimization problems.     

A computationally efficient evolutionary algorithm for real-parameter optimization

Due to increasing interest in solving real-world optimization problems using evolutionary algorithms (EAs), researchers have recently developed a number of real-parameter genetic algorithms (GAs). In these studies, the main research effort is spent on developing an efficient recombination operator. Such recombination operators use probability distributions around the parent solutions to create an offspring. Some operators emphasize solutions at the center of mass of parents and some around the parents. In this paper, we propose a generic parent-centric recombination operator (PCX) and a steady-state, elite-preserving, scalable, and computationally fast population-alteration model (we call the G3 model). The performance of the G3 model with the PCX operator is investigated on three commonly used test problems and is compared with a number of evolutionary and classical optimization algorithms including other real-parameter GAs with the unimodal normal distribution crossover (UNDX) and the simplex crossover (SPX) operators, the correlated self-adaptive evolution strategy, the covariance matrix adaptation evolution strategy (CMA-ES), the differential evolution technique, and the quasi-Newton method. The proposed approach is found to consistently and reliably perform better than all other methods used in the study. A scale-up study with problem sizes up to 500 variables shows a polynomial computational complexity of the proposed approach. This extensive study clearly demonstrates the power of the proposed technique in tackling real-parameter optimization problems.     

A graph-based parameterization concept for global laminate optimization

A new graph-based parameterization concept aimed at the global optimization of laminated structures by the means of evolutionary algorithms and finite element analysis is introduced. The motivation to develop this novel parameterization concept is twofold. First, the entire design space is accessible to optimi zation down to the smallest entity, which is a single finite element, and secondly, this concept guarantees greatest flexibility in terms of laminate layer shape and placement. The finite element mesh of a structure is represented as a mathematical graph. Substructures of this graph form fiber reinforced and possibly overlapping patches and are affiliated to virtual graph vertices representing their properties. Adapted genetic variation operators are directly applied on this graph. The method allows for concurrent optimization of number, size, shape, and position of the patches and an arbitrary number of material related properties for each of them. The novel concept overcomes the limits of traditional geometry-based approaches, as it is able to represent almost arbitrary patch shapes even on curved surfaces. Two numerical examples demonstrate the efficiency of the method.     

Multiobjective optimization using population-based extremal optimization

In recent years, a general-purpose local-search heuristic method called Extremal Optimization (EO) has been successfully applied in some NP-hard combinatorial optimization problems. In this paper, we present a novel Pareto-based algorithm, which can be regarded as an extension of EO, to solve multiobjective optimization problems. The proposed method, called Multiobjective Population-based Extremal Optimization (MOPEO), is validated by using five benchmark functions and metrics taken from the standard literature on multiobjective evolutionary optimization. The experimental results demonstrate that MOPEO is competitive with the state-of-the-art multiobjective evolutionary algorithms. Thus MOPEO can be considered as a viable alternative to solve multiobjective optimization problems.     

Tuning compilations by multi-objective optimization: Application to Apache web server

Modern compilers present a great and ever increasing number of options which can modify the features and behavior of a compiled program. Many of these options are often wasted due to the required comprehensive knowledge about both the underlying architecture and the internal processes of the compiler. In this context, it is usual, not having a single design goal but a more complex set of objectives. In addition, the dependencies between different goals are difficult to be a priori inferred. This paper proposes a strategy for tuning the compilation of any given application. This is accomplished by using an automatic variation of the compilation options by means of multi-objective optimization and evolutionary computation commanded by the NSGA-II algorithm. This allows finding compilation options that simultaneously optimize different objectives. The advantages of our proposal are illustrated by means of a case study based on the well-known Apache web server. Our strategy has demonstrated an ability to find improvements up to 7.5% and up to 27% in context switches and L2 cache misses, respectively, and also discovers the most important bottlenecks involved in the application performance.     

Learning algorithm for multimodal optimization

We present a new evolutionary algorithm—“learning algorithm” for multimodal optimization. The scheme for reproducing a new generation is very simple. Control parameters, of the length of the list of historical best solutions and the “learning probability” of the current solutions being moved towards the current best solutions and towards the historical ones, are used to assign different search intensities to different parts of the feasible area and to direct the updating of the current solutions. Results of numerical tests on minimization of the 2D Schaffer function, the 2D Shubert function and the 10D Ackley function show that this algorithm is effective and efficient in finding multiple global solutions of multimodal optimization problems.     

Adaptive population-based search: Application to estimation of nonlinear regression parameters

Algorithms for the estimation of nonlinear regression parameters are considered. Adaptive population-based search algorithms are proposed and implemented in deriving reliable estimates at a reasonable time with default setting of their controlling parameters. The algorithms are tested on the NIST collection of data sets containing 27 nonlinear regression tasks of various level of difficulty. The experimental results show that both algorithms with competing heuristics are significantly more reliable as compared with the algorithm based on Levenberg-Marquardt optimizing procedure.     

An evolutionary method for complex-process optimization

In this paper we present a new evolutionary method for complex-process optimization. It is partially based on the principles of the scatter search methodology, but it makes use of innovative strategies to be more effective in the context of complex-process optimization using a small number of tuning parameters. In particular, we introduce a new combination method based on path relinking, which considers a broader area around the population members than previous combination methods. We also use a population-update method which improves the balance between intensification and diversification. New strategies to intensify the search and to escape from suboptimal solutions are also presented. The application of the proposed evolutionary algorithm to different sets of both state-of-the-art continuous global optimization and complex-process optimization problems reveals that it is robust and efficient for the type of problems intended to solve, outperforming the results obtained with other methods found in the literature.     

A hierarchical neighbourhood search method for topology optimization

This paper presents a hierarchical neighbourhood search method for solving topology optimization problems defined on discretized linearly elastic continuum structures. The design of the structure is represented by binary design variables indicating material or void in the various finite elements.Two different designs are called neighbours if they differ in only one single element, in which one of them has material while the other has void. The proposed neighbourhood search method repeatedly jumps to the best neighbour of the current design until a local optimum has been found, where no further improvement can be made. The engine of the method is an efficient exploitation of the fact that if only one element is changed (from material to void or from void to material) then the new global stiffness matrix is just a low-rank modification of the old one. To further speed up the process, the method is implemented in a hierarchical way. Starting from a coarse finite element mesh, the neighbourhood search is repeatedly applied on finer and finer meshes.Numerical results are presented for minimum-weight problems with constraints on respectively compliance, strain energy densities in all non-void elements, and von Mises stresses in all non-void elements.     

Superior solution guided particle swarm optimization combined with local search techniques

Particle swarm optimization (PSO) is an evolutionary algorithm known for its simplicity and effectiveness in solving various optimization problems. PSO should have strong yet balanced exploration and exploitation capabilities to enhance its performance. A superior solution guided PSO (SSG-PSO) framework integrated with an individual level based mutation operator and different local search techniques is proposed in this study. In SSG-PSO, a collection of superior solutions is maintained and updated with the evolutionary process, such that each particle can comprehensively learn from the recorded superior solutions. In addition, to maintain the diversity of the particle swarm, SSG-PSO is combined with an individual level based mutation operator, which will be invoked when a particle is trapped in a local optimum (determined by the fitness and position states of the particle), thereby improving the adaptation and flexibility of each individual particle. Moreover, two gradient-based local search techniques, namely, the Broyden–Fletcher–Goldfarb–Shanno (BFGS) and Davidon–Fletcher–Powell (DFP) Quasi–Newton methods, and two derivative-free local search techniques, namely, pattern search and Nelder–Mead simplex search, are incorporated into SSG-PSO. The performances of SSG-PSO and that of its local search enhanced variants are extensively and comparatively studied on a suit of benchmark optimization functions.     

A hybrid particle swarm optimization algorithm for high-dimensional problems

In recent years, particle swarm optimization (PSO) emerges as a new optimization scheme that has attracted substantial research interest due to its simplicity and efficiency. However, when applied to high-dimensional problems, PSO suffers from premature convergence problem which results in a low optimization precision or even failure. To remedy this fault, this paper proposes a novel memetic PSO (CGPSO) algorithm which combines the canonical PSO with a Chaotic and Gaussian local search procedure. In the initial evolution phase, CGPSO explores a wide search space that helps avoid premature convergence through Chaotic local search. Then in the following run phase, CGPSO refines the solutions through Gaussian optimization. To evaluate the effectiveness and efficiency of the CGPSO algorithm, thirteen high dimensional non-linear scalable benchmark functions were examined. Results show that, compared to the standard PSO, CGPSO is more effective, faster to converge, and less sensitive to the function dimensions. The CGPSO was also compared with two PSO variants, CPSO-H, DMS-L-PSO, and two memetic optimizers, DEachSPX and MA-S2. CGPSO is able to generate a better, or at least comparable, performance in terms of optimization accuracy. So it can be safely concluded that the proposed CGPSO is an efficient optimization scheme for solving high-dimensional problems.     

Structure optimization of neural networks for evolutionary design optimization

We study the use of neural networks as approximate models for the fitness evaluation in evolutionary design optimization. To improve the quality of the neural network models, structure optimization of these networks is performed with respect to two different criteria: One is the commonly used approximation error with respect to all available data, and the other is the ability of the networks to learn different problems of a common class of problems fast and with high accuracy. Simulation results from turbine blade optimizations using the structurally optimized neural network models are presented to show that the performance of the models can be improved significantly through structure optimization.We would like to thank the BMBF, grant LOKI, number 01 IB 001 C, for their financial support of our research.     

Global Simplex Optimization—A simple and efficient metaheuristic for continuous optimization

A new hybrid optimization algorithm is proposed for minimization of continuous multi-modal functions. The algorithm called Global Simplex Optimization (GSO) is a population set based Evolutionary Algorithm (EA) incorporating a special multi-stage, stochastic and weighted version of the reflection operator of the classical simplex method. An optional mutation operator has also been tested and then removed from the structure of the final algorithm in favor of simplicity and because of insignificant effect on performance. The promising performance achieved by GSO is demonstrated by comparisons made to some other state-of-the-art global optimization algorithms over a set of conventional benchmark problems.     

A novel group search optimizer for multi-objective optimization

In this paper, a novel multi-objective group search optimizer named NMGSO is proposed for solving the multi-objective optimization problems. To simplify the computation, the scanning strategy of the original GSO is replaced by the limited pattern search procedure. To enrich the search behavior of the rangers, a special mutation with a controlling probability is designed to balance the exploration and exploitation at different searching stages and randomness is introduced in determining the coefficients of members to enhance the diversity. To handle multiple objectives, the non-dominated sorting scheme and multiple producers are used in the algorithm. In addition, the kernel density estimator is used to keep diversity. Simulation results based on a set of benchmark functions and comparisons with some methods demonstrate the effectiveness and robustness of the proposed algorithm, especially for the high-dimensional problems.     

A dominance tree and its application in evolutionary multi-objective optimization

Most contemporary multi-objective evolutionary algorithms (MOEAs) store and handle a population with a linear list, and this may impose high computational complexities on the comparisons of solutions and the fitness assignment processes. This paper presents a data structure for storing the whole population and their dominating information in MOEAs. This structure, called a Dominance Tree (DT), is a binary tree that can effectively and efficiently store three-valued relations (namely dominating, dominated or non-dominated) among vector values. This paper further demonstrates DT’s potential applications in evolutionary multi-objective optimization with two cases. The first case utilizes the DT to improve NSGA-II as a fitness assignment strategy. The second case demonstrates a DT-based MOEA (called a DTEA), which is designed by leveraging the favorable properties of the DT. The simulation results show that the DT-improved NSGA-II is significantly faster than NSGA-II. Meanwhile, DTEA is much faster than SPEA2, NSGA-II and an improved version of NSGA-II. On the other hand, in regard to converging to the Pareto optimal front and maintaining the diversity of solutions, DT-improved NSGA-II and DTEA are found to be competitive with NSGA-II and SPEA2.     

An improved vector particle swarm optimization for constrained optimization problems

Increasing attention is being paid to solve constrained optimization problems (COP) frequently encountered in real-world applications. In this paper, an improved vector particle swarm optimization (IVPSO) algorithm is proposed to solve COPs. The constraint-handling technique is based on the simple constraint-preserving method. Velocity and position of each particle, as well as the corresponding changes, are all expressed as vectors in order to present the optimization procedure in a more intuitively comprehensible manner. The NVPSO algorithm [30], which uses one-dimensional search approaches to find a new feasible position on the flying trajectory of the particle when it escapes from the feasible region, has been proposed to solve COP. Experimental results showed that searching only on the flying trajectory for a feasible position influenced the diversity of the swarm and thus reduced the global search capability of the NVPSO algorithm. In order to avoid neglecting any worthy position in the feasible region and improve the optimization efficiency, a multi-dimensional search algorithm is proposed to search within a local region for a new feasible position. The local region is composed of all dimensions of the escaped particle’s parent and the current positions. Obviously, the flying trajectory of the particle is also included in this local region. The new position is not only present in the feasible region but also has a better fitness value in this local region. The performance of IVPSO is tested on 13 well-known benchmark functions. Experimental results prove that the proposed IVPSO algorithm is simple, competitive and stable.     

A parallel double-level multiobjective evolutionary algorithm for robust optimization

Robust optimization is a popular method to tackle uncertain optimization problems. However, traditional robust optimization can only find a single solution in one run which is not flexible enough for decision-makers to select a satisfying solution according to their preferences. Besides, traditional robust optimization often takes a large number of Monte Carlo simulations to get a numeric solution, which is quite time-consuming. To address these problems, this paper proposes a parallel double-level multiobjective evolutionary algorithm (PDL-MOEA). In PDL-MOEA, a single-objective uncertain optimization problem is translated into a bi-objective one by conserving the expectation and the variance as two objectives, so that the algorithm can provide decision-makers with a group of solutions with different stabilities. Further, a parallel evolutionary mechanism based on message passing interface (MPI) is proposed to parallel the algorithm. The parallel mechanism adopts a double-level design, i.e., global level and sub-problem level. The global level acts as a master, which maintains the global population information. At the sub-problem level, the optimization problem is decomposed into a set of sub-problems which can be solved in parallel, thus reducing the computation time. Experimental results show that PDL-MOEA generally outperforms several state-of-the-art serial/parallel MOEAs in terms of accuracy, efficiency, and scalability.     

A restarted and modified simplex search for unconstrained optimization

In this paper we propose a simple but efficient modification of the well-known Nelder–Mead (NM) simplex search method for unconstrained optimization. Instead of moving all n simplex vertices at once in the direction of the best vertex, our “shrink” step moves them in the same direction but one by one until an improvement is obtained. In addition, for solving non-convex problems, we simply restart the so-modified NM (MNM) method by constructing an initial simplex around the solution obtained in the previous phase. We repeat restarts until there is no improvement in the objective function value. Thus, our restarted modified NM (RMNM) is a descent and deterministic method and may be seen as an extended local search for continuous optimization. In order to improve computational complexity and efficiency, we use the heap data structure for storing and updating simplex vertices. Extensive empirical analysis shows that: our modified method outperforms in average the original version as well as some other recent successful modifications; in solving global optimization problems, it is comparable with the state-of-the-art heuristics.     

A heuristic iterated-subspace minimization method with pattern search for unconstrained optimization

Recently, an increasing attention was paid on different procedures for an unconstrained optimization problem when the information of the first derivatives is unavailable or unreliable. In this paper, we consider a heuristic iterated-subspace minimization method with pattern search for solving such unconstrained optimization problems. The proposed method is designed to reduce the total number of function evaluations for the implementation of high-dimensional problems. Meanwhile, it keeps the advantages of general pattern search algorithm, i.e., the information of the derivatives is not needed. At each major iteration of such a method, a low-dimensional manifold, the iterated subspace, is constructed. And an approximate minimizer of the objective function in this manifold is then determined by a pattern search method. Numerical results on some classic test examples are given to show the efficiency of the proposed method in comparison with a conventional pattern search method and a derivative-free method.