Showing papers in "Evolutionary Programming in 1998"
TL;DR: This paper first analyzes the impact that inertia weight and maximum velocity have on the performance of the particle swarm optimizer, and then provides guidelines for selecting these two parameters.
Abstract: This paper first analyzes the impact that inertia weight and maximum velocity have on the performance of the particle swarm optimizer, and then provides guidelines for selecting these two parameters. Analysis of experiments demonstrates the validity of these guidelines.
3,557 citations
TL;DR: This paper compares two evolutionary computation paradigms: genetic algorithms and particle swarm optimization, and suggests ways in which performance might be improved by incorporating features from one paradigm into the other.
Abstract: This paper compares two evolutionary computation paradigms: genetic algorithms and particle swarm optimization. The operators of each paradigm are reviewed, focusing on how each affects search behavior in the problem space. The goals of the paper are to provide additional insights into how each paradigm works, and to suggest ways in which performance might be improved by incorporating features from one paradigm into the other.
1,661 citations
TL;DR: This paper investigates the philosophical and performance differences of particle swarm and evolutionary optimization by comparison experiments involving four non-linear functions well studied in the evolutionary optimization literature.
Abstract: This paper investigates the philosophical and performance differences of particle swarm and evolutionary optimization. The method of processing employed in each technique are first reviewed followed by a summary of their philosophical differences. Comparison experiments involving four non-linear functions well studied in the evolutionary optimization literature are used to highlight some performance differences between the techniques.
1,163 citations
TL;DR: A simplified version of the particle swarm algorithm is examined in an effort to understand the trajectories of particles as they search for solutions and address optimal parameter values.
Abstract: The particle swarm algorithm has been shown to optimize a wide variety of complex functions. This paper examines a simplified version of the algorithm in an effort to understand the trajectories of particles as they search for solutions. Findings address optimal parameter values, point out issues for future research, and contribute to understanding this new optimization method.
350 citations
TL;DR: It is shown that evolutionary algorithms are able to converge to the set of minimal elements in finite time with probability one, provided that the search space is finite, the time-invariant variation operator is associated with a positive transition probability function and that the selection operator obeys the so-called ‘elite preservation strategy.’
Abstract: The task of finding minimal elements of a partially ordered set is a generalization of the task of finding the global minimum of a real-valued function or of finding Pareto-optimal points of a multicriteria optimization problem. It is shown that evolutionary algorithms are able to converge to the set of minimal elements in finite time with probability one, provided that the search space is finite, the time-invariant variation operator is associated with a positive transition probability function and that the selection operator obeys the so-called ‘elite preservation strategy.’
82 citations
TL;DR: A new method for the fully automated and rapid flexible docking of inhibitors covalently bound to serine proteases is described, which combines an energy function specifically tuned for molecular docking and an evolutionary programming search engine and takes advantage of the constained geometry about the site of covalent attachment to dramatically limit the search space and increase search efficiency.
Abstract: Viral serine proteases have become increasingly attractive targets for rational drug design. Many known inhibitors of serine proteases form a covalent bond to the activated serine oxygen, an interaction not taken into account by available docking software used for database mining. We describe a new method for the fully automated and rapid flexible docking of inhibitors covalently bound to serine proteases. The method combines an energy function specifically tuned for molecular docking and an evolutionary programming search engine, and takes advantage of the constained geometry about the site of covalent attachment to dramatically limit the search space and increase search efficiency. Results for several test systems are presented, including a database search which yielded a known inhibitor as a highranking compound.
60 citations
TL;DR: The results show that the adaptive EA outperforms the other two approaches to Boolean satisfiability problems, and suggests that this EA is not only a good solver for satisfiable problems, but for constraint satisfaction problems in general.
Abstract: We investigate three approaches to Boolean satisfiability problems. We study and compare the best heuristic algorithm WGSAT and two evolutionary algorithms, an evolution strategy and an evolutionary algorithm adapting its own fitness function while running. The results show that the adaptive EA outperforms the other two approaches. The power of this EA originates from the adaptive mechanism, which is completely problem independent and generally applicable to any constraint satisfaction problem. This suggests that the adaptive EA is not only a good solver for satisfiability problems, but for constraint satisfaction problems in general.
58 citations
TL;DR: This paper investigates empirically how well the self-adaptation scheme of EP works on a set of benchmark functions and an experimental evaluation of an existing simple fix to the problem is carried out.
Abstract: Evolutionary programming (EP) has been widely used in numerical optimization in recent years. One of EP's key features is its self-adaptation scheme. In EP, mutation is typically the only operator used to generate new offspring. The mutation is often implemented by adding a random number from a certain distribution (e.g., Gaussian in the case of classical EP) to the parent. An important parameter of the Gaussian distribution is its standard deviation (or equivalently the variance). In the widely used self-adaptation scheme of EP, this parameter is evolved, rather than manually fixed, along with the objective variables. This paper investigates empirically how well the self-adaptation scheme works on a set of benchmark functions. Some anomalies have been observed in the empirical studies, which demonstrate that the self-adaptation scheme may not work as well as hoped for some functions. An experimental evaluation of an existing simple fix to the problem is also carried out in this paper.
40 citations
TL;DR: This paper reports work investigating various evolutionary approaches to vertex cover (VC), a well-known NP-Hard optimization problem, that incorporates features of a powerful traditional heuristic for VC that allow initial evolutionary algorithm (EA) populations to be seeded in known promising regions of the search space.
Abstract: This paper reports work investigating various evolutionary approaches to vertex cover (VC), a well-known NP-Hard optimization problem. Central to each of the algorithms is a novel encoding scheme for VC and related problems that treats each chromosome as a binary decision diagram. As a result, the encoding allows only a (guaranteed optimal) subset of feasible solutions. The encoding also incorporates features of a powerful traditional heuristic for VC that allow initial evolutionary algorithm (EA) populations to be seeded in known promising regions of the search space. The resulting EAs have displayed exceptionally strong empirical performance on various vertex cover, independent set, and maximum clique problem classes.
37 citations
TL;DR: It is shown that the sometimes invoked model of a perturbed gradient search does not seem to give an appropriate picture of the search process, and the search behavior is described as the antagonism of exploitation and exploration, where exploitation works in one dimension, whereas the exploration is a random walk on a (N−1)-dimensional manifold in the search space.
Abstract: This paper discusses the question how ES/EP-like algorithms perform the evolutionary search in real-valued N-dimensional parameter spaces It will be shown that the sometimes invoked model of a perturbed gradient search does not seem to give an appropriate picture of the search process Instead, the search behavior is described as the antagonism of exploitation and exploration, where exploitation works in one dimension, whereas the exploration is a random walk on a (N−1)-dimensional manifold in the search space As an example the exploration dynamics on the sphere model will be investigated
37 citations
TL;DR: A dynamic fitness function that is founded in previous analysis done on both static and dynamic landscapes, and that avoids problematic issues with previously proposed dynamic landscapes for GAs is proposed.
Abstract: We attempt to find mutation / crossover rate pairs that facilitate the performance of a genetic algorithm (GA) on a simple dynamic fitness function. This research results in two products. The first is a dynamic fitness function that is founded in previous analysis done on both static and dynamic landscapes, and that avoids problematic issues with previously proposed dynamic landscapes for GAs. The second is a general relationship between the crossover and mutation rates that are most useful for a dynamic fitness function with a specific rate of change in Hamming distance, and that could possibly provide insight into the utility of the standard GA approach for the optimization of dynamic landscapes.
TL;DR: It is shown that mutation can do as well as, or better than, recombination even under low inter-agent epistasis; fitness sharing is shown to alter the characteristics of the coevolving system.
Abstract: This paper examines a key aspect of applying evolutionary computing techniques to multi-agent systems: a comparison in the performance of the genetic operators of mutation and recombination Using the tuneable NKC model of multi-agent evolution it is shown that the benefits of simple recombination and mutation vary depending on the type of system, with bit mutation capable of doing as well as recombination in systems with significant inter-agent epistasis The effects of fitness sharing between the interacting individuals are then examined and it is shown that mutation can do as well as, or better than, recombination even under low inter-agent epistasis; fitness sharing is shown to alter the characteristics of the coevolving system
TL;DR: This paper investigates the behaviour of four different EP algorithms for large-scale problems, i.e., problems whose dimension ranges from 100 to 300, and identifies classical EP (CEP), fast EP (FEP), which are classical and fast EP respectively.
Abstract: Most analytical and experimental results on evolutionary programming (EP) are obtained using low-dimensional problems, e.g., smaller than 50. It is unclear, however, whether the empirical results obtained from the low-dimensional problems still hold for high-dimensional cases. This paper investigates the behaviour of four different EP algorithms for large-scale problems, i.e., problems whose dimension ranges from 100 to 300. The four are classical EP (CEP) [1, 2], fast EP (FEP).
TL;DR: The main purpose is to find an objective method to evaluate strategies for the Classical Iterated Prisoner's Dilemma, and shows, with a genetic approach, how basic ideas could be used in order to generate automatically a great numbers of strategies.
Abstract: The Classical Iterated Prisoner's Dilemma (CIPD) is used to study the evolution of cooperation. We show, with a genetic approach, how basic ideas could be used in order to generate automatically a great numbers of strategies. Then we show some results of ecological evolution on those strategies, with the description of the experimentations we have made. Our main purpose is to find an objective method to evaluate strategies for the CIPD. Finally we use the former results to add a new argument confirming that there is, in order to be good, an infinite gradient in the level of complexity in structure of strategies.
TL;DR: It is found statistically significant appearance of general adaptive features in a spatially distributed population of prisoner's dilemma playing agents in a noisy environment.
Abstract: We investigate the following question. Do populations of evolving agents adapt only to their recent environment or do general adaptive features appear over time? We find statistically significant appearance of general adaptive features in a spatially distributed population of prisoner's dilemma playing agents in a noisy environment. Multiple populations are evolved in an evolutionary algorithm structured as a cellular automaton with states drawn from a rich set of prisoner's dilemma strategies. Populations are sampled early and at the end of a ten-thousand generation simulation. Modern and archaic populations are then placed in competition. We test the hypothesis that competition between an archaic and modern population yields probability p=0.5 of modern populations out-competing archaic ones. The hypothesis is rejected at a confidence level of 99.5% using a binomial probability model in each of seven variations of our basic experiment.
TL;DR: In this paper, an agent-based computational economics (ACE) model of a labor market with choice and refusal of contractual partners and endogenously evolving work-site behavior is presented, where experimentally determined correlations between market structure and the formation and evolution of contractual networks, and between contractual network formation and the types of worksite interactions and social welfare outcomes.
Abstract: This paper reports on computational experiments for an agent-based computational economics (ACE) model of a labor market with choice and refusal of contractual partners and endogenously evolving work-site behavior. Three types of labor market structures are examined: two-sided markets comprising workers and employers; partially fluid markets comprising pure workers, pure employers, and agents capable of functioning both as workers and as employers; and endogenous-type markets in which each agent is capable of functioning as both a worker and an employer. Particular attention is focused on experimentally determined correlations between market structure and the formation and evolution of contractual networks, and between contractual network formation and the types of work-site interactions and social welfare outcomes that these contractual networks support.
TL;DR: An evolutionary optimization method based on a hybrid of an interior penalty and augmented Lagrangian function ensures the generation of feasible solutions during the evolutionary search process with less computation time than required by the interior method.
Abstract: An evolutionary optimization method based on a hybrid penalty function is proposed for the general constrained optimization problem. As an extension of the earlier method of Evolian (evolutionary optimization based on Lagrangian), the difference comes in the form of the penalty function. The hybrid of an interior penalty and augmented Lagrangian function ensures the generation of feasible solutions during the evolutionary search process with less computation time than required by the interior method. Some numerical results indicate the effectiveness of the hybrid penalty method on several optimization problems.
TL;DR: Simulation results indicate that when the number of elements in the array is greater than 120, EP is always successful in finding solutions that have maximum sidelobe levels of −20 dB or lower.
Abstract: Thinning of large antenna arrays in order to obtain low sidelobes is a difficult and nontrivial problem. Evolutionary programming (EP), a multi-agent stochastic search method, is proposed for optimizing thinned phased arrays with a large number of elements. Experiments are conducted to determine the efficiency of EP in terms of the reliability in producing acceptable solutions and the quality of the final solution. Antenna arrays with up to 200 elements are thinned to obtain maximum sidelobe levels of less than −20 dB. Simulation results indicate that when the number of elements in the array is greater than 120, EP is always successful in finding solutions that have maximum sidelobe levels of −20 dB or lower.
TL;DR: Empirical evidence is provided that this method provides a statistically significant improvement in GP system performance for symbolic regression problems, and GP runs are more likely to find a solution, and successful runs use fewer generations.
Abstract: A weakness of genetic programming (GP) is the difficulty it suffers in discovering useful numeric constants for the terminal nodes of the s-expression trees. We examine a solution to this problem, called numeric mutation, based, roughly, on simulated annealing. We provide empirical evidence to demonstrate that this method provides a statistically significant improvement in GP system performance for symbolic regression problems. GP runs are more likely to find a solution, and successful runs use fewer generations.
TL;DR: The analysis presented here indicates, in contrast with previous literature, that the introduction of noise to the evaluation of solutions can change the expected sampling of schemata, even when the noise is zero mean.
Abstract: Traditional selection in genetic algorithms has relied on reproduction in proportion to observed fitness. There has been recent interest in assessing the result of proportional selection on schemata in the presence of random effects (e.g., noisy evaluation of solutions). The analysis presented here indicates, in contrast with previous literature, that the introduction of noise to the evaluation of solutions can change the expected sampling of schemata, even when the noise is zero mean. Unfortunately, this “misallocation of trials” can also result simply from random initialization of a population.
TL;DR: A novel approach to the use of automatically defined functions (ADFs) with the help of a genetic library builder (GLiB) to automatically create subpopulations of ADFs during evolution, where these are termed evolution-defined functions (EDFs).
Abstract: In this paper we introduce a novel approach to the use of automatically defined functions (ADFs) with the help of a genetic library builder (GLiB). The new technique uses the two mutation operators of GLiB to automatically create subpopulations of ADFs during evolution, where these are termed evolution-defined functions (EDFs). Our approach consists of dynamically specifying separate subpopulations for each identified ADF, where a further population of programs uses individuals from these subpopulations during evaluations. Using a multiplexer problem and two classification tasks we compare a number of existing methods with this co-evolutionary approach. It is shown that dynamically creating ADF subpopulations (according to worth) proves more beneficial than specifying them a priori. It is also shown that the approach performs better than existing approaches — GP with ADFs and GP with GLiB — at all three tasks. Further, we extend the approach to allow the number of EDFs to emerge during the course of evolution, removing the need to specify how many functions are available a priori.
TL;DR: The experiments indicate that, in the test environment, expert knowledge is best incorporated only in the initial population, which is a welcome result as this is the computationally inexpensive choice of the two methods of incorporating expert knowledge tested.
Abstract: We present a family of related test problems for genetic programming. These test problems form a very simple test environment that nevertheless possesses some degree of algorithmic subtlety. We term this genetic programming environment plus-one-recall-store (PORS). This genetic programming environment has only a pair of terminals, 1 and recall, and a pair of operations, plus and store, together with a single memory location. We present an extensive mathematical characterization of the PORS environment and report experiments testing the benefits of incorporating expert knowledge into the initial population and into the operation of crossover. The experiments indicate that, in the test environment, expert knowledge is best incorporated only in the initial population. This is a welcome result as this is the computationally inexpensive choice of the two methods of incorporating expert knowledge tested.
TL;DR: This paper discusses one particular class of boundary operators — sphere operators — and discusses their applicability to some constrained problems (with convex feasible search spaces) through a mapping between the boundary of the feasible region of the search space and a sphere.
Abstract: In this paper we continue our earlier studies [13, 14] on boundary operators for constrained parameter optimization problems. The significance of this line of research is based on the observation that usually the global solution for many optimization problems lies on the boundary of the feasible region. Thus, for many constrained numerical optimization problems it might be beneficial to search just the boundary of the search space defined by a set of constraints (some other algorithm might be used for searching the interior of the search space, if activity of a constraint is not certain). We discuss one particular class of boundary operators — sphere operators — and discuss their applicability to some constrained problems (with convex feasible search spaces) through a mapping between the boundary of the feasible region of the search space and a sphere. We provide also with some experimental evaluation of these transformations.
TL;DR: Simulation results indicate that the EP procedure is capable of selecting appropriate basis functions for different regions of the input space as well as optimizing the associated set of parameters.
Abstract: The idea of evolutionary friendliness recognizes that problem representations have a significant impact on the performance of evolutionary algorithms. There are two aspects of these representations. Different solution schemes exploit different natural symmetries. Very commonly, problems also possess symmetries that are determined by the coordinate systems used to represent them. Solution symmetries are typically specified by the user and are not allowed to evolve. The problem coordinate system is again typically chosen by the user and not evolved. In this first paper, the most appropriate solution symmetry is evolved. In the second paper, the coordinate system is evolved. In this paper, common detection problems with decision boundaries that possess special symmetries are solved using an evolutionary programming (EP) framework that is capable of exploiting these symmetries to quickly generate solutions. In particular, neural networks possessing appropriate symmetries are evolved by optimizing both their bases and their parameters. Simulation results indicate that the EP procedure is capable of selecting appropriate basis functions for different regions of the input space as well as optimizing the associated set of parameters.
TL;DR: It is demonstrated that a solution for artificial ant problem is very likely to be non-general and relying on the specific characteristics of the Santa Fe trail, and the method can be useful in producing general behaviours for simulation environments.
Abstract: This research aims to demonstrate that a solution for artificial ant problem [4] is very likely to be non-general and relying on the specific characteristics of the Santa Fe trail. It then presents a consistent method which promotes producing general solutions. Using the concepts of training and testing from machine learning research, the method can be useful in producing general behaviours for simulation environments.
TL;DR: A visualization program is presented together with a number of examples that illustrate some aspects of evolution of phenotypes in unlimited, multidimensional, real spaces under different kinds of selection.
Abstract: Decades after the potential of evolutionary inspirations in global optimization was acknowledged, there is still some confusion concerning the way the essential evolutionary operators act in basic adaptive situations. This is because the dynamics of evolving populations are quite complex and often counterintuitive. Therefore appropriate visualization programs might be really helpful in providing the required expertise. We present such a program together with a number of examples that illustrate some aspects of evolution of phenotypes in unlimited, multidimensional, real spaces under different kinds of selection.
TL;DR: This approach starts with a traditional AI planner and uses GP to acquire control rules to improve its efficiency and two ways to introduce domain knowledge acquired by another method (Hamlet) into EvoCK: seeding the initial population and using a new operator (knowledge-based crossover).
Abstract: In this paper we describe EvoCK, a new approach to the application of genetic programming (GP) to planning This approach starts with a traditional AI planner (Prodigy) and uses GP to acquire control rules to improve its efficiency We also analyze two ways to introduce domain knowledge acquired by another method (Hamlet) into EvoCK: seeding the initial population and using a new operator (knowledge-based crossover) This operator combines genetic material from both an evolving population and a non-evolving population containing background knowledge We tested these ideas in the blocksworld domain and obtained excellent results
TL;DR: A steady state memetic algorithm is shown to be successful in matching shapes even when they are partially obscured, and even in the presence of noise in the input image.
Abstract: Shape matching techniques are important in machine intelligence, especially in applications such as robotics. Currently, there are three major approaches to shape recognition: statistical, syntactic and neural approaches. This paper presents a fourth approach: evolutionary algorithms. A steady state memetic algorithm is shown to be successful in matching shapes even when they are partially obscured, and even in the presence of noise in the input image.
TL;DR: Using evolutionary programming, this work has simulated the binding of probes of length four nucleotides to a series of target lengths to determine most optimal target length that can be unambiguously reconstructed.
Abstract: DNA sequencing methods are the subject of continued interest in molecular biology for use in a wide variety of applications. Sequencing DNA by hybridization on a “DNA chip” has been estimated to increase the rate of DNA sequencing by as much as one-million fold. In this process, the sequence of a target molecule is reconstructed by the complementary binding of a pool of random probe molecules. For each target, an appropriate probe length must be used to unambiguously determine the sequence of a given target sequence of length N. Using evolutionary programming, we have simulated the binding of probes of length four nucleotides to a series of target lengths to determine most optimal target length that can be unambiguously reconstructed. Evolutionary programming is demonstrated to be well suited to sequence reconstruction problems and could also be extended for gene expression monitoring with DNA chip technology.
TL;DR: A two-way algorithm that was evolved using genetic programming with 83% accuracy for determining whether a protein is an extracellular protein, 84% for nuclear proteins, 89% for membrane proteins, and 83% for anchored membrane proteins is described.
Abstract: As newly sequenced proteins are deposited into the world's ever-growing archives, they are typically immediately tested by various algorithms for clues as to their biological structure and function. One question about a new protein involves its cellular location — that is, where the protein resides in a living organism (e.g., extracellular, membrane, nuclear). A human-created five-way algorithm for cellular location using statistical techniques with 76% accuracy was recently reported. This paper describes a two-way algorithm that was evolved using genetic programming with 83% accuracy for determining whether a protein is an extracellular protein, 84% for nuclear proteins, 89% for membrane proteins, and 83% for anchored membrane proteins. Unlike the statistical calculation, the genetically evolved programs employ a large and varied arsenal of computational capabilities, including arithmetic functions, conditional operations, subroutines, iterations, named memory, indexed memory, setcreating operations, and look-ahead. The genetically evolved classification program can be viewed as an extension (which we call a programmatic motif) of the conventional notion of a protein motif.