Showing papers on "Multi-swarm optimization published in 1995"
TL;DR: A snapshot of particle swarming from the authors’ perspective, including variations in the algorithm, current and ongoing research, applications and open problems, is included.
Abstract: A concept for the optimization of nonlinear functions using particle swarm methodology is introduced The evolution of several paradigms is outlined, and an implementation of one of the paradigms is discussed Benchmark testing of the paradigm is described, and applications, including nonlinear function optimization and neural network training, are proposed The relationships between particle swarm optimization and both artificial life and genetic algorithms are described
18,439 citations
04 Oct 1995
TL;DR: The optimization of nonlinear functions using particle swarm methodology is described and implementations of two paradigms are discussed and compared, including a recently developed locally oriented paradigm.
Abstract: The optimization of nonlinear functions using particle swarm methodology is described. Implementations of two paradigms are discussed and compared, including a recently developed locally oriented paradigm. Benchmark testing of both paradigms is described, and applications, including neural network training and robot task learning, are proposed. Relationships between particle swarm optimization and both artificial life and evolutionary computation are reviewed.
14,477 citations
TL;DR: This paper presents a distributed genetic algorithm for optimization of large structures on a cluster of workstations connected via a local area network (LAN) based on its adaptability to a high degree of parallelism.
Abstract: Parallel algorithms for optimization of structures reported in the literature have been restricted to shared-memory multiprocessors. This paper presents a distributed genetic algorithm for optimization of large structures on a cluster of workstations connected via a local area network (LAN). The selection of genetic algorithm is based on its adaptability to a high degree of parallelism. Two different approaches are used to transform the constrained structural optimization problem to an unconstrained optimization problem: a penalty-function method and augmented Lagrangian approach. For the solution of the resulting simultaneous linear equations the iterative preconditioned conjugate gradient (PCG) method is used because of its low memory requirement. A dynamic load-balancing mechanism is developed to account for the unpredictable multiuser, multasking environment of a networked cluster of workstations, heterogeneity of machines, and indeterminate nature of the interative PCG equation solver. The algorithm ...
188 citations
TL;DR: From the computational results, it can conclude that the large-step optimization methods outperform the simulated annealing method and find more frequently an optimal schedule than the other studied methods.
133 citations
19 Jun 1995
TL;DR: In this article, an aerodynamic shape optimization method based on a genetic algorithm with the compressible Euler or Navier-Stokes equations is proposed to find optimum aerodynamic configurations efficiently.
Abstract: This paper presents an aerodynamic shape optimization method based on a genetic algorithm with the compressible Euler or Navier-Stokes equations. A new algorithm to represent aerodynamic configurations is proposed to find optimum aerodynamic configurations efficiently. We solve three aerodynamic shape optimization problems. First, an inverse problem is solved to understand a basic character and to examine the possibility of the genetic algorithm. Second, a drag minimization problem is solved to find a streamlined body. Finally, L/D(ratio of lift to drag) maximization problem in a transonic flow is solved and the optimized configuration is compared with a supercritical airfoil. The results show that genetic algorithm is powerful and robust in solving aerodynamic shape optimization problems.
52 citations
[...]
19 citations
14 citations
20 Mar 1995
TL;DR: Fuzzy rules, which describe the human supervision during the optimization process, are combined with the numerical rules of the original algorithm to refine the output of each iteration.
Abstract: The optimization process can be viewed as a closed-loop control system. Traditional "controllers", the numerical optimization algorithms, are usually "crisply" designed for well defined mathematical models. However, when applied to engineering design optimization problems in which function evaluations can be expensive and imprecise, very often the crisp algorithms will become impractical or will not converge. A common strategy for designers is to monitor the optimization process and keep "tuning" it in an interactive manner, using their judgment on the information obtained from the previous iterations, and their knowledge of the problem. This paper presents how the heuristics of this human supervision can be modeled into the optimization algorithms using fuzzy control concept. A fuzzy version of sequential linear programming, which is very popular in engineering design optimization, is used to demonstrate this idea. Fuzzy rules, which describe the human supervision during the optimization process, are combined with the numerical rules of the original algorithm to refine the output of each iteration. Simple numerical examples are used to show the feasibility and practicality of this approach. >
13 citations
TL;DR: An optimization algorithm, based on Creutz's microcanonical simulation technique, which has proven very efficient for non-convex optimization tasks associated with image-processing applications and should also constitute a useful heuristic for applications in other domains requiring combinatorial optimization searches.
12 citations
Proceedings Article•
11 Sep 1995
TL;DR: An adaptive optimizer with two features is proposed and design that features a “learning” capability for canned queries that allows existing plans to be incrementally replaced by “fitter” ones.
Abstract: Traditionally, optimizers are “programmed” to optimize queries following a set of buildin procedures. However, optimizers should be robust to its changing environment to generate the fittest query execution plans. To realize adaptiveness, we propose and design an adaptive optimizer with two features. First, the search space and search strategy of the optimizer can be tuned by parameters to allow the optimizer to pick the one that fits best during the optimization process. Second, the optimizer features a “learning” capability for canned queries that allows existing plans to be incrementally replaced by “fitter” ones. An experimental study on large multijoin queries based on an analytical model is used to demonstrate the effectiveness of such an approach.
12 citations
29 Nov 1995
TL;DR: Some powerful direct parameter optimization algorithms are presented, being combinations of direct global and local search methods, and the basic structure of an optimization strategy is described, able to accomplish an extensive analysis of the optimum points of a given cost function.
Abstract: Today, a great shortcoming of the existing direct global optimization methods like Genetic Algorithms, Evolution Strategies, Simulated Annealing, etc. is, that they are only approximation algorithms usually requiring high numbers of cost function evaluations. Hence, in case of cost functions which are expensive to evaluate, these algorithms are not applicable any more. In this paper, some powerful direct parameter optimization algorithms are presented, being combinations of direct global and local search methods. Beyond that, the basic structure of an optimization strategy is described, which is able to accomplish an extensive analysis of the optimum points of a given cost function (multiple-stage optimization). Our developed methods are implemented and integrated into REM0 (REsearch Model Optimization package) 19, 111 representing a software tool for experimentation and optimization of simulation models. Some optimization results are presented to demonstrate that our approach successfully focuses the advantages of global and local search.
Proceedings Article•
15 Jul 1995
TL;DR: In this article, an algorithm for approximating solutions of multicriterial nonlinear optimization problems of a class of discontinuous objective functions with disconnected constraint set is proposed in the framework of integral based optimization theory under some weak general assumptions.
Abstract: An algorithm for approximating solutions of multicriterial nonlinear optimization problems of a class of discontinuous objective functions with disconnected constraint set is proposed in the framework of integral based optimization theory under some weak general assumptions. Numerical examples are given to illustrate the effectiveness of the algorithm
TL;DR: In this paper, a class of combinatorial optimization problems with several criteria of the form MINMAX and MINSUM in arbitrary combinations is investigated, and polynomial algorithms are proposed which find a complete set of alternatives for the problems on paths, matchings, spanning trees, and for the integer transportation problem.
Abstract: A class of combinatorial optimization problems with several criteria of the form MINMAX and MINSUM in arbitrary combinations is investigated. A class of vector optimization problems is formulated in terms of systems of subsets. It contains optimization problems on graphs and Boolean programming. It is proved that finding a complete set of alternatives for such problems can be of exponential complexity. New results concerning the existence of statistically efficient algorithms for these problems are obtained. For the case of two criteria (when at least one of them is MINMAX) polynomial algorithms are proposed which find a complete set of alternatives for the problems on paths, matchings, spanning trees, and for the integer transportation problem. This work was partially supported by the Byelorussian Foundation for Fundamental Research. 1. DEFINITIONS AND NOTATIONS Following the terminology of [1], a system of subsets (SS) is a pair Γ = (Ε, Τ), where Ε is a finite set and Τ is a family of non-empty subsets of the set Ε which are called trajectories. For the set / G T the number \\t\\ is called the length of trajectory. For the set E, we define a vector weight function (VWF) W(e) = (u;1(e),.. . ,u; r(e))> where iu , (e)eR + , se #r = {l,2,...,r}, e£E. We consider an r-criterial combinatorial problem (r-CCP) such that for the set of trajectories Τ {t} a vector cost function (VCF) is defined, and we wish to minimize F8(t) for all 5 G Nr: Fs(t) -> min Vs e Nr , where the criteria may be of the forms MINMAX: Fs(t) = max w3(e) — > min, (1.1) MINSUM: F,(t) = ws(t) = £ w,(e) -> min (1.2) e6i in arbitrary combinations. Such problem will be denoted by Z =(T,W,F) = (E,T,W,F). *UDC 519.1. Originally published in Diskretnaya Matematika (1995) 7, No. 1, 3-18 (in Russian). Translated by N. N. Kuzyurin. 94 V. A. Emelichev and M K. Kravtsov The problem Z will be called minimax (minisum) if the VCF F(t) consists of the criteria MINMAX (MINSUM). It is obvious that a lot of multi-criterial graph problems (on spanning trees, the travelling salesman problem, on perfect matchings, on paths, etc.), which were considered in [2-10], are special cases of the above r-CCP. As usual, a trajectory t G T is called Pareto-optimal if there is no trajectory t' e T such that F(t') < F(t) and F(t') ^ F(t). By f we denote the Pareto set (PS) of all Pareto-optimal trajectories of the problem Z. Recall that the complete set of alternatives £CSA) is defined as a subset f C f of the minimal size which satisfies the equality F(f) = F(f\\ where F(T') = {F(t): t e T} VT' C T. By an r-CCP, the problem of finding and representing in explicit form the PS f or the CSA T is meant, where the r-CCP is a mass problem [1,11,12]. In such problem a form of some criteria and a type of trajectories with some properties may be fixed (for example, in terms of graphs that may be circuits, spanning trees, matchings, etc.). Along with the mass problem we use the notion of individual problem. Any individual r-CCP is defined by a specific triple (Γ, W, F), where the SS Γ = (Ε, Τ) and all parameters of the VCF F(t), including the VWF W(e), are fixed. In this case the SS Γ is called an individual VWF. Now we give a list of discrete problems which are considered in this paper recalling in each case the sense of a set of feasible solutions X = {x} (in the sequel we show that all these problems may be considered also as r-CCPs with corresponding VCFs): Z\\ is the travelling salesman problem, X is the set of Hamiltonian circuits of a graph; Z2 is the spanning tree problem, X is the set of spanning trees of a graph; Z3 is the path problem, X is the set of simple paths between two given vertices of a graph; Z^ is the matching problem, X is the set of perfect matchings of a graph with even number of vertices; Z5 is the assignment problem [13] of order m χ η, m < n, X = {z = \\\\xij\\\\mxn. Xij E {0,1} V(i, j) G Nm χ 7Vn, £ xtj = l Vi G Nm, E χ» < l Vj e Nn\\ ; j=l t=l ) Z6 is the assignment problem of order m x n, m < n, with m types of jobs [14], Χ = {χ = H x t f l l m x n : X i j Ε {0,1} V( i , j ) G Nm χ 7Vn, n m -^ E Xij = at V» G 7Vm, E Xij < l Vj G 7Vn k j=l i=l J where at is the number of jobs of iu\\ type, αϊ + ... + am = n; On combinatorial vector optimization problems 95 Z7 is the distribution problem with Boolean variables [15, 16] of order πι χ η, Χ = x = Mmxnl Xtj G {Ο, 1} V(t, j) G Μη Χ Nn, τη η Ν Σ zy = ι Vj e Nn, £ yxy < o, v» e wm ; i=l j=l J is the one-parametrical standardization problem [17] of order m x n, m > η, Χ = ζ = lli tf lUx»: ζϋ 6 {0, 1} V(i, j) e Λ» x ΛΓη, m χ Σ *y = l W e 7Vn, iy = 0 Vi < j ; i = l ' is the p-median graph problem [18], χ = z = iKIUx«: χϋ e {0, 1}
TL;DR: In this article, the authors proposed a new approach which is totally different from the forementioned conventional methods, i.e., the global optimization method for the unconstrained nonlinear optimization problem where chaos is introduced.
Abstract: Most of the actual optimization problems are nonlinear and have many peaks (nonconvex). With the widespread use of the high-speed and large-capacity computers as the background, it has recently been felt highly necessary to derive the global optimal solution for the optimization problem which is nonlinear and has multiple peaks. It is one of the most important topics of research in the field of optimization.
The major global optimization method developed until now can be divided into the trajectory method, the function transformation method, and the simulated annealing method.
This paper proposes a new approach which is totally different from the forementioned conventional methods, i.e., the global optimization method for the unconstrained nonlinear optimization problem where chaos is introduced. the proposed method is based on the dissipative system where the inertia term and the nonlinear damping term are added to the conventional gradient method.
By appropriately adjusting the characteristics of the nonlinear damping term, the generation of chaos can itself be controlled. Then, by overriding the barrier of the energy function between the local minima, the process converges to the global optimal solution.
Finally, the proposed method is applied to the typical multipeaked nonlinear optimization problem with two and ten variables and it is shown that the global optimal solution can be derived by adequately adjusting the parameter of the nonlinear damping term.
14 Nov 1995
TL;DR: The techniques for quantitatively measuring the performance of population based multiobjective optimization algorithms and techniques for automatically designing optimization algorithms are developed and demonstrated.
Abstract: In this paper, we develop and demonstrate techniques to design multiobjective optimization algorithms based on hybrid fuzzy system/evolutionary algorithm techniques. The technique is based on an approach where a fuzzy system is used to control the evolutionary algorithm. By viewing the search process as a dynamic process, high performance strategies are developed using controller design techniques. Through the use of indicators aimed at assessing the performance of evolutionary algorithms for multiobjective optimization, we show how to design fuzzy systems for controlling the search behavior. The key contributions of the work reported in this paper are the techniques for quantitatively measuring the performance of population based multiobjective optimization algorithms and techniques for automatically designing optimization algorithms. We demonstrate our techniques on an Integrated Circuit placement task that includes timing and geometrical objectives.
01 Jan 1995
TL;DR: Applied to the signal decomposition problem of RBF approximation, the behaviour of the adiabatic layering model is shown to be in close correspondence with the theoretical expectations.
TL;DR: This note discusses the performances and applications of two methods generally used in structural optimization, one is the direct method which applies a nonlinear programming (NLP) algorithm directly to the structural optimization problem and the other is the approximation method which utilizes the engineering sense very well.
Abstract: This note discusses the performances and applications of two methods generally used in structural optimization. One is the direct method which applies a nonlinear programming (NLP) algorithm directly to the structural optimization problem. The other is the approximation method which utilizes the engineering sense very well. The two methods are compared through standard structural optimization problems with truss and beam elements. The results are analysed based on the convergence performances, the number of function calculations, the quality of the cost functions, etc. The applications of both methods are also discussed.
TL;DR: It is shown that any lexicographic optimum in vector optimization problems on a finite set of admissible solutions can be obtained by a classical technique, the linear convolution of criteria.
Abstract: It is shown that any lexicographic optimum in vector optimization problems on a finite set of admissible solutions can be obtained by a classical technique, the linear convolution of criteria.
01 Jan 1995
TL;DR: The work on flow control and optimization at the Air Force Center for Optimal Design and Control (CODAC) has focused on new tools for improved aerodynamic design and the optimal and feedback control of fluid flows.
Abstract: The work on flow control and optimization at the Air Force Center for Optimal Design and Control (CODAC) has focused on new tools for improved aerodynamic design and the optimal and feedback control of fluid flows.
22 Oct 1995
TL;DR: A hybridization of accelerated evolutionary programming and a deterministic optimization procedure is applied to a series of constrained nonlinear and quadratic optimization problems and results indicate that the hybrid approach can outperform the other methods when addressing constrained optimization problems.
Abstract: A hybridization of accelerated evolutionary programming (AEP) and a deterministic optimization procedure is applied to a series of constrained nonlinear and quadratic optimization problems. The hybrid scheme is compared with other existing schemes such as AEP alone, two-phase (TP) optimization, and EP with a nonstationary penalty function (NS-EP). The results indicate that the hybrid approach can outperform the other methods when addressing constrained optimization problems with respect to the computational efficiency and solution accuracy.
13 Dec 1995
TL;DR: This paper considers certain features of optimization-based design, wherein the physical laws, expressed as boundary value problems, are incorporated as constraints, and a modification of the optimization algorithm to include a linearized constraint which effectively reduces the dimensionality.
Abstract: In this paper we consider certain features of optimization-based design, wherein the physical laws, expressed as boundary value problems, are incorporated as constraints. One such idea is a modification of the optimization algorithm to include a linearized constraint which effectively reduces the dimensionality. A second approach exploits domain decomposition ideas to analyze the physics on separate domains with compatibility being enforced by the optimization algorithm. Both cases suggest computational approaches that provide significant savings over methods that do not exploit problem structure.
21 Jun 1995
TL;DR: In this article, the joint problem of optimization and parameter estimation of a nonlinear dynamic continuous reactor is considered, where decision analysis is incorporated into the integrated optimization and parametric estimation scheme in order to reflect conflicting situations derived from the simultaneous consideration of multiple performance criteria.
Abstract: The integrated problem of optimization and parameter estimation is addressed in the context of vector-valued optimization techniques. Some elements of decision analysis are incorporated into the integrated optimization and parameter estimation scheme in order to reflect conflicting situations derived from the simultaneous consideration of multiple performance criteria. To illustrate the main aspects of the proposed approach, the joint problem of optimization and parameter estimation of a nonlinear dynamic continuous reactor is considered. Decentralization and concurrency of the acquisition, identification and optimization tasks are also considered.