Showing papers presented at "Congress on Evolutionary Computation in 2003"

Data clustering using particle swarm optimization

Abstract: This paper proposes two new approaches to using PSO to cluster data. It is shown how PSO can be used to find the centroids of a user specified number of clusters. The algorithm is then extended to use K-means clustering to seed the initial swarm. This second algorithm basically uses PSO to refine the clusters formed by K-means. The new PSO algorithms are evaluated on six data sets, and compared to the performance of K-means clustering. Results show that both PSO clustering techniques have much potential.

Pareto-based multi-objective differential evolution

Abstract: Evolutionary multiobjective optimization (EMOO) finds a set of Pareto solutions rather than any single aggregated optimal solution for a multiobjective problem. The purpose is to describe a newly developed evolutionary approach-Pareto-based multiobjective differential evolution (MODE). The concept of differential evolution, which is well-known in the continuous single-objective domain for its fast convergence and adaptive parameter setting, is extended to the multiobjective problem domain. A Pareto-based approach is proposed to implement the differential vectors. A set of benchmark test functions is used to validate this new approach. We compare the computational results with those obtained in the literature, specifically by strength Pareto evolutionary algorithm (SPEA). It is shown that this new approach tends to be more effective in finding the Pareto front in the sense of accuracy and approximate representation of the real Pareto front with comparable efficiency.

Artificial immune system (AIS) research in the last five years

Abstract: Immunity-based techniques are gaining popularity in wide area of applications, and emerging as a new branch of artificial intelligence (AI). The paper surveys the major works in this field during the last five years, in particular, it reviews the works of existing methods and the new initiatives.

A critical survey of performance indices for multi-objective optimisation

Abstract: A large number of methods for solving multiobjective optimisation (MOO) problems have been developed. To compare these methods rigorously, or to measure the performance of a particular MOO algorithm quantitatively, a variety of performance indices (PIs) have been proposed. We provide an overview of the various PIs and attempts to categorise them into a certain number of classes according to their properties. Comparative studies have been conducted using a group of artificial solution sets and a group of solution sets obtained by various MOO solvers to show the advantages and disadvantages of the PIs. The comparative studies show that many PIs may be misleading in that they fail to truly reflect the quality of solution sets. Thus, it may not be a good practice to evaluate the performance of MOO solvers based on PIs only.

Differential evolution for multi-objective optimization

Abstract: Two test problems on multiobjective optimization (one simple general problem and the second one on an engineering application of cantilever design problem) are solved using differential evolution (DE). DE is a population based search algorithm, which is an improved version of genetic algorithm (GA), Simulations carried out involved solving (1) both the problems using Penalty function method, and (2) first problem using Weighing factor method and finding Pareto optimum set for the chosen problem, DE found to be robust and faster in optimization. To consolidate the power of DE, the classical Himmelblau function, with bounds on variables, is also solved using both DE and GA. DE found to give the exact optimum value within less generations compared to simple GA.

Particle swarm optimization with mutation

Abstract: The particle swarm optimization algorithms converges rapidly during the initial stages of a search, but often slows considerably and can get trapped in local optima. This paper examines the use of mutation to both speed up convergence and escape local minima. It compares the effectiveness of the basic particle swarm optimization scheme (BPSO) with each of BPSO with mutation, constriction particle swarm optimization (CPSO) with mutation, and CPSO without mutation. The four test functions used were the Sphere, Ackley, Rastrigin and Rosenbrock functions of dimensions 10, 20 and 30. The results show that mutation hinders the motion of the swarm on the sphere but the combination of CPSO with mutation provides a significant improvement in performance for the Rastrigin and Rosenbrock functions for all dimensions and the Ackley function for dimensions 20 and 30, with no improvement for the 10 dimensional case.

Evolutionary many-objective optimisation: an exploratory analysis

Abstract: This inquiry explores the effectiveness of a class of modern evolutionary algorithms, represented by NSGA-II, for solving optimisation tasks with many conflicting objectives Optimiser behaviour is assessed for a grid of recombination operator configurations Performance maps are obtained for the dual aims of proximity to, and distribution across, the optimal trade-off surface Classical settings for recombination are shown to be suitable for small numbers of objectives but correspond to very poor performance as the number of objectives is increased, even when large population sizes are used Explanations for this behaviour are offered via the concepts of dominance resistance and active diversity promotion

Multiple single objective Pareto sampling

Abstract: We detail a new nonPareto evolutionary multiobjective algorithm, multiple single objective Pareto sampling (MSOPS), that performs a parallel search of multiple conventional target vector based optimisations, e.g. weighted min-max. The method can be used to generate the Pareto set and analyse problems with large numbers of objectives. The method allows bounds and discontinuities of the Pareto set to be identified and the shape of the surface to be analysed, despite not being able to visualise the surface easily. A new combination metric is also introduced that allows the shape of the objective surface that gives rise to discontinuities in the Pareto surface to be analysed easily.

Neuroevolution for reinforcement learning using evolution strategies

Abstract: We apply the CMA-ES, an evolution strategy which efficiently adapts the covariance matrix of the mutation distribution, to the optimization of the weights of neural networks for solving reinforcement learning problems. It turns out that the topology of the networks considerably influences the time to find a suitable control strategy. Still, our results with fixed network topologies are significantly better than those reported for the best evolutionary method so far, which adapts both the weights and the structure of the networks.

A first study of fuzzy cognitive maps learning using particle swarm optimization

Abstract: We introduce a new algorithm for fuzzy cognitive maps learning. The proposed approach is based on the particle swarm optimization method and it is used for the detection of proper weight matrices that lead the fuzzy cognitive map to desired steady states. For this purpose a properly defined objective function that incorporates experts' knowledge is constructed and minimized. The application of the proposed methodology to an industrial control problem supports the claim that the proposed technique is efficient and robust.

Empirical study of particle swarm optimizer with an increasing inertia weight

Abstract: A PSO with increasing inertia weight, distinct from a widely used PSO with decreasing inertia weight, is proposed in this paper. Four standard test functions with asymmetric initial range settings are used to confirm its validity. From the experiments, it is clear that a PSO with increasing inertia weight outperforms the one with decreasing inertia weight, both in convergent speed and solution precision, with no additional computing load compared with the PSO with a decreasing inertia weight.

Non-stationary problem optimization using the primal-dual genetic algorithm

Abstract: Genetic algorithms (GAs) have been widely used for stationary optimization problems where the fitness landscape does not change during the computation. However, the environments of real world problems may change over time, which puts forward serious challenge to traditional GAs. In this paper, we introduce the application of a new variation of GA called the primal-dual genetic algorithm (PDGA) for problem optimization in nonstationary environments. Inspired by the complementarity and dominance mechanisms in nature, PDGA operates on a pair of chromosomes that are primal-dual to each other in the sense of maximum distance in genotype in a given distance space. This paper investigates an important aspect of PDGA, its adaptability to dynamic environments. A set of dynamic problems are generated from a set of stationary benchmark problems using a dynamic problem generating technique proposed in this paper. Experimental study over these dynamic problems suggests that PDGA can solve complex dynamic problems more efficiently than traditional GA and a peer GA, the dual genetic algorithm. The experimental results show that PDGA has strong viability and robustness in dynamic environments.

Web usage mining using artificial ant colony clustering and linear genetic programming

Abstract: The rapid e-commerce growth has made both business community and customers face a new situation. Due to intense competition on the one hand and the customer's option to choose from several alternatives, the business community has realized the necessity of intelligent marketing strategies and relationship management. Web usage mining attempts to discover useful knowledge from the secondary data obtained from the interactions of the users with the Web. Web usage mining has become very critical for effective Web site management, creating adaptive Web sites, business and support services, personalization, network traffic flow analysis and so on. The study of ant colonies behavior and their self-organizing capabilities is of interest to knowledge retrieval/management and decision support systems sciences, because it provides models of distributed adaptive organization, which are useful to solve difficult optimization, classification, and distributed control problems, among others [Ramos, V. et al. (2002), (2000)]. In this paper, we propose an ant clustering algorithm to discover Web usage patterns (data clusters) and a linear genetic programming approach to analyze the visitor trends. Empirical results clearly show that ant colony clustering performs well when compared to a self-organizing map (for clustering Web usage patterns) even though the performance accuracy is not that efficient when compared to evolutionary-fuzzy clustering (i-miner) [Abraham, A. (2003)] approach.

A reputation-based trust model for peer-to-peer e-commerce communities

Abstract: Peer-to-peer e-commerce (electronic commerce) communities are commonly perceived as an environment offering both opportunities and threats. One way to minimize threats in such an open community is to use community-based reputations to help evaluating the trustworthiness and predicting the future behavior of peers. We present PeerTrust - a coherent adaptive trust model for quantifying and comparing the trustworthiness of peers based on a transaction-based feedback system. There are two main features of our model. First, we introduce three basic trust parameters in computing trustworthiness of peers. In addition to feedback a peer receives through its transactions with other peers, we incorporate the total number of transactions a peer performs, and the credibility of the feedback sources into the model for evaluating the trustworthiness of peers. We argue that the trust models based solely on feedback from other peers in the community are inaccurate and ineffective. Second, we introduce two adaptive trust factors, the transaction context factor and the community context factor, to allow the basic trust metric to incorporate different contexts (situations) and to address common problems encountered in a variety of online e-commerce communities. We present a concrete method to validate the proposed trust model and report the set of simulation-based experiments, showing the feasibility and benefit of the PeerTrust model.

On the use of particle swarm optimization with multimodal functions

Abstract: We present two hybrid particle swarm optimization (PSO) algorithms that incorporate a mutation operator similar to the one used with evolutionary algorithms. We study our hybridized PSO algorithm with two schemes called g/spl I.bar/best and l/spl I.bar/best, and we apply them to multimodal functions. The proposed approaches are validated using test functions taken from the specialized literature, and our results are compared with respect to those obtained by other highly competitive PSO algorithms. Our comparative study indicates that the hybridization of PSO with a nonuniform mutation operator significantly improves its performance when dealing with multimodal functions.

Parameter identification of induction motors using differential evolution

Abstract: Parameter identification of system models is a fundamental step in the process of designing a controller for a system. In control engineering, a wide selection of analytic identification techniques exists for linear systems, but not for nonlinear systems. Instead, the model parameters may be determined by an optimization algorithm by minimizing the error between model output and measured data. We apply the differential evolution algorithm to parameter identification of two induction motors. The motors are used in the house circulation pumps produced by the Danish pump manufacturer Grundfos A/S. The experiments presented use differential evolution, and is a follow-up study of an comparison of eight stochastic search algorithms on the two motor identification problems. In conclusion, the differential evolution algorithm outperformed the previously best known algorithms on both problems.

Evolution strategies assisted by Gaussian processes with improved preselection criterion

Abstract: In many engineering optimization problems, the number of fitness function evaluations is limited by time and cost. These problems pose a special challenge to the field of evolutionary computation, since existing evolutionary methods require a very large number of problem function evaluations. One popular way to address this challenge is the application of approximation models as a surrogate of the real fitness function. We propose a model assisted evolution strategy, which uses a Gaussian process approximation model to preselect the most promising solutions. To refine the preselection process we determine the likelihood of each individual to improve the overall best found solution. Due to this, the new algorithm has a much better convergence behavior and achieves better results than standard evolutionary optimization approaches with less fitness evaluations. Numerical results from extensive simulations on several high dimensional test functions including multimodal functions are presented.

A comparison of constraint-handling methods for the application of particle swarm optimization to constrained nonlinear optimization problems

Abstract: We present a comparison of two constraint-handling methods used in the application of particle swarm optimization (PSO) to constrained nonlinear optimization problems (CNOPs). A brief review of constraint-handling techniques for evolutionary algorithms (EAs) is given, followed by a direct comparison of two existing methods of enforcing constraints using PSO. The two methods considered are the application of nonstationary multistage penalty functions and the preservation of feasible solutions. Five benchmark functions are used for the comparison, and the results are examined to assess the performance of each method in terms of accuracy and rate of convergence. Conclusions are drawn and suggestions for the applicability of each method to real-world CNOPs are given.

Bounded archiving using the lebesgue measure

Abstract: Many modern multiobjective evolutionary algorithms (MOEAs) store the points discovered during optimization in an external archive, separate from the main population, as a source of innovation and/or for presentation at the end of a run. Maintaining a bound on the size of the archive may be desirable or necessary for several reasons, but choosing which points to discard and which to keep in the archive, as they are discovered, is not trivial. We briefly review the state-of-the-art in bounded archiving, and present a new method based on locally maximizing the hyper-volume dominated by the archive. The new archiver is shown to outperform existing methods, on several problem instances, with respect to the quality of the archive obtained when judged using three distinct quality measures.

Evolving adaptive neural networks with and without adaptive synapses

Abstract: A potentially powerful application of evolutionary computation (EC) is to evolve neural networks for automated control tasks. However, in such tasks environments can be unpredictable and fixed control policies may fail when conditions suddenly change. Thus, there is a need to evolve neural networks that can adapt, i.e. change their control policy dynamically as conditions change. In this paper, we examine two methods for evolving neural networks with dynamic policies. The first method evolves recurrent neural networks with fixed connection weights, relying on internal state changes to lead to changes in behavior. The second method evolves local rules that govern connection weight changes. The surprising experimental result is that the former method can be more effective than evolving networks with dynamic weights, calling into question the intuitive notion that networks with dynamic synapses are necessary for evolving solutions to adaptive tasks.

Elitist multiobjective evolutionary algorithm for environmental/economic dispatch

Abstract: The environmental/economic dispatch problem is a multiobjective nonlinear optimization problem with constraints. Until recently, this problem has been addressed by considering economic and emission objectives separately or as a weighted sum of both objectives. Multiobjective evolutionary algorithms can find multiple Pareto-optimal solutions in one single run and this ability makes them attractive for solving problems with multiple and conflicting objectives. We use an elitist multiobjective evolutionary algorithm based on the nondominated sorting genetic algorithm-II (NSGA-II) for solving the environmental/economic dispatch problem. Elitism ensures that the population best solution does not deteriorate in the next generations. Simulation results are presented for a sample power system.

Pareto neuro-evolution: constructing ensemble of neural networks using multi-objective optimization

Abstract: In this paper, we present a comparison between two multiobjective formulations to the formation of neuro-ensembles The first formulation splits the training set into two nonoverlapping stratified subsets and form an objective to minimize the training error on each subset, while the second formulation adds random noise to the training set to form a second objective A variation of the memetic Pareto artificial neural network (MPANN) algorithm is used MPANN is based on differential evolution for continuous optimization The ensemble is formed from all networks on the Pareto frontier It is found that the first formulation outperformed the second The first formulation is also found to be competitive to other methods in the literature

AISEC: an artificial immune system for e-mail classification

Abstract: With the increase in information on the Internet, the strive to find more effective tools for distinguishing between interesting and non-interesting material is increasing. Drawing analogies from the biological immune system, this paper presents an immune-inspired algorithm called AISEC that is capable of continuously classifying electronic mail as interesting and non-interesting without the need for re-training. Comparisons are drawn with a naive Bayesian classifier and it is shown that the proposed system performs as well as the naive Bayesian system and has a great potential for augmentation.

The role of /spl epsi/-dominance in multi objective particle swarm optimization methods

Abstract: In this paper, the influence of /spl epsi/-dominance on multi-objective particle swarm optimization (MOPSO) methods is studied. The most important role of /spl epsi/-dominance is to bound the number of non-dominated solutions stored in the archive (archive size), which has influences on computational time, convergence and diversity of solutions. Here, /spl epsi/-dominance is compared with the existing clustering technique for fixing the archive size and the solutions are compared in terms of computational time, convergence and diversity. A new diversity metric is also suggested. The results show that the /spl epsi/-dominance method can find solutions much faster than the clustering technique with comparable and even in some cases better convergence and diversity.

Expected runtimes of a simple multi-objective evolutionary algorithm

Abstract: The expected runtime of a simple multi-objective evolutionary algorithm for the Boolean decision space is analyzed. The algorithm uses independent bit flips as mutation operator and, therefore, searches globally. It is proved that the expected runtime is O(n/sup n/) for all objective functions {0,1}/sup n/ /spl rarr/ R/sup m/. This worst-case bound is tight and matches the worst-case bounds for fundamental evolutionary algorithms working in the scenario of single-objective optimization. For the bicriteria problem LOTZ (leading ones trailing zeroes), it is shown that the expected runtime is O(n/sup 3/). Moreover, the runtime is O(n/sup 3/) with an overwhelming probability.

An evolution strategy with probabilistic mutation for multi-objective optimisation

Abstract: Evolutionary algorithms have been applied with great success to the difficult field of multiobjective optimisation. Nevertheless, the need for improvements in this field is still strong. We present a new evolutionary algorithm, ESP (the Evolution Strategy with Probabilistic mutation). ESP extends traditional evolution strategies in two principal ways: it applies mutation probabilistically in a GA-like fashion, and it uses a new hyper-volume based, parameterless, scaling independent measure for resolving ties during the selection process. ESP outperforms the state-of-the-art algorithms on a suite of benchmark multiobjective test functions using a range of popular metrics.

Hybrid evolutionary algorithms based on PSO and GA

Abstract: Inspired by the idea of genetic algorithm, we propose two hybrid evolutionary algorithms based on PSO and GA methods through crossing over the PSO and GA algorithms. The main ideas of the two proposed methods are to integrate PSO and GA methods in parallel and series forms respectively. Simulations for a series of benchmark test functions show that both of the two proposed methods possess better ability to find the global optimum than that of the standard PSO algorithm.

Beyond "Web of trust": enabling P2P e-commerce

Abstract: The huge success of eBay has proven the demand for customer-to-customer (C2C) electronic commerce. eBay is a centralized infrastructure with all its scalability problems (network bandwidth, server load, availability, etc.). We argue that C2C e-commerce is an application domain that maps naturally onto the emergent field of P2P systems simply by its underlying interaction model of customers, i.e., peers. This offers the opportunity to take P2P systems beyond mere file sharing systems into interesting new application domains. The long-term goal would be to design a fully functional decentralized system which resembles eBay without eBay's dedicated, centralized infrastructure. Since security (authenticity, non-repudiation, trust, etc.) is key to any e-commerce infrastructure, our envisioned P2P e-commerce platform has to address this adequately. As the first step in this direction we present an approach for a completely decentralized P2P public key infrastructure (PKI) which can serve as the basis for higher-level security service. In contrast to other systems in this area, such as PGP which uses a "Web of trust" concept, we use a statistical approach which allows us to provide an analytical model with provable guarantees, and quantify the behavior and specific properties of the PKI. To justify our claims we provide a first-order analysis and discuss its resilience against various known threats and attack scenarios. In support of our belief that C2C E-commerce is one of the potential killer applications of the emerging structured P2P systems, we provide a layered model for P2P E-commerce, demonstrating the dependencies of various security related issues that can be built on top of a decentralized PKI.

AntTree: a new model for clustering with artificial ants

Abstract: We present a new clustering algorithm for unsupervised learning. It is inspired from the self-assembling behavior observed in real ants where ants progressively become attached to an existing support and then successively to other attached ants. The artificial ants that we have defined similarly builds a tree. Each ant represents one data. The way ants move and build this tree depends on the similarity between the data. We have compared our results to those obtained by the k-means algorithm and by AntClass on numerical databases (either artificial, real, or from the CE.R.I.E.S.). We show that AntTree significantly improves the clustering process.

Secure supply-chain protocols

Abstract: Supply chain interactions have huge economic importance, yet these interactions are managed inefficiently. One of the major sources of inefficiency in supply-chain management is information asymmetry; i.e., information that is available to one or more organizations in the chain (e.g., manufacturer, retailer) is not available to others. There are several causes of information asymmetry, among them fear that a powerful buyer or supplier will take advantage of private information, that information will leak to a competitor, etc. We propose secure supply-chain collaboration (SSCC) protocols that enable supply-chain partners to cooperatively achieve desired system-wide goals without revealing the private information of any of the parties, even though the jointly computed decisions require the information of all the parties. Secure supply-chain collaboration has the potential to improve supply-chain management practice, and by removing a major inefficiency therein, improves productivity. We present specific SSCC protocols for two types of supply-chain interactions: capacity allocation, and e-auctions (electronic auctions). In the course of doing so, we design techniques that are of independent interest, and are likely to be useful in the design of future SSCC protocols.