evolutionary algorithms (EAs) are populationbased, stochastic search algorithms that mimic natural evolution. Over the years, EAs have been successfully applied to many classification problems. In this paper, we prese...
详细信息
evolutionary algorithms (EAs) are populationbased, stochastic search algorithms that mimic natural evolution. Over the years, EAs have been successfully applied to many classification problems. In this paper, we present three novel evolutionary approaches and analyze their performances for synthesizing classifiers with EAs in supervised data mining scenarios. The first approach is based on encoding rule sets with bit string genomes, while the second one utilizes Genetic Programming (GP) to create decision trees with arbitrary expressions attached to the nodes. The novelty of these two approaches lies in the use of solutions on the Pareto front as an ensemble. The third approach, EDDIE-101, is also based on GP but uses a new, advanced fitness measure and some novel genetic operators. We compare these approaches to a number of well-known data mining methods, including C4.5 and Random-Forest, and show that the performances of our evolved classifiers can be very competitive as far as the solution quality is concerned. In addition, the proposed approaches work well across a wide range of configurations, and EDDIE-101 particularly has been highly efficient. To further evaluate the flexibility of EDDIE-101 across different problem domains, we also test it on some real financial datasets for finding investment opportunities and compare the results with those obtained using other classifiers. Numerical experiments confirm that EDDIE-101 can be successfully extended to financial forecasting.
There exist several methods for binary classification of gene expression data sets. However, in the majority of published methods, little effort has been made to minimize classifier complexity. In view of the small nu...
详细信息
There exist several methods for binary classification of gene expression data sets. However, in the majority of published methods, little effort has been made to minimize classifier complexity. In view of the small number of samples available in most gene expression data sets, there is a strong motivation for minimizing the number of free parameters that must be fitted to the data. In this paper, a method is introduced for evolving (using an evolutionary algorithm) simple classifiers involving a minimal subset of the available genes. The classifiers obtained by this method perform well, reaching 97% correct classification of clinical outcome on training samples from the breast cancer data set published by van't Veer, and up to 89% correct classification on validation samples from the same data set, easily outperforming previously published results.
The calibration of a hydrological model is an important task for obtaining accurate runoff simulation results for a specific watershed. Several optimisation algorithms have been applied during the last years for the a...
详细信息
The calibration of a hydrological model is an important task for obtaining accurate runoff simulation results for a specific watershed. Several optimisation algorithms have been applied during the last years for the automatic calibration of conceptual rainfall-runoff (CRR) models. The aim of this study is to compare the effectiveness and the efficiency of three evolutionary algorithms, namely the Shuffled Complex Evolution (SCE), the Genetic algorithms (GA) and the evolutionary Annealing-Simplex (EAS), for the calibration of the Medbasin-D daily CRR model. An improved calibration approach of Medbasin-D is presented, including a batch-processing module which enables the implementation of coupled simulation-optimisation routines. The enhanced Medbasin calibration module is employed in a watershed of the island of Crete (Greece), using several objective functions in order to test the optimisation algorithms under different hydrological flow conditions. The results reveal that, in terms of effectiveness, SCE and EAS performed equally well, while GA provided slightly worse optimal solutions. However, GA was computationally more efficient than SCE and EAS. Despite the discrepancies among the optimisation runs, the simulated hydrographs had a very similar response for the optimal parameter sets obtained by the same calibration criteria, indicating that all tested optimisation methods produce equally successful results with Medbasin-D model. Additionally, the selected objective function seems to have a more decisive effect on the final simulation outcomes.
Population diversity is very important in giving the algorithm the power to explore the search space and not get trapped in local optima. In this respect, using a probabilistic representation for the quantum individua...
详细信息
Population diversity is very important in giving the algorithm the power to explore the search space and not get trapped in local optima. In this respect, using a probabilistic representation for the quantum individuals, the Quantum-inspired evolutionary algorithms (QiEA) claim higher diversity in the population. Here, considering this important feature of QiEA, we propose different structures to offer better interaction between the q- individuals and propose new operators to preserve the diversity in the population and thus improve the performance of the QiEA. The effect of the structured population is investigated on the performance of the algorithm. Additionally, two operators are proposed in this paper. Being called the Diversity Preserving QiEA the first operator finds the converged similar q- individuals around a local optimum and while keeping the best q- individuals, by reinitializing the inferior ones pushes them out of the basin of attraction of the local optimum, so helping the algorithm to search other regions in the search space. The other operator is a reinitialization operator which by reinitializing the whole population helps it escape from the local optima it is trapped in. By studying the effect of the parameters of the proposed operators on their performance we show how the proposed operators improve the performance of QiEA. Experiments are performed on Knapsack, Trap and fourteen numerical objective functions and the results show better performance for the proposed algorithm than the original version of QiEA.
Warehouse location problems play a significant role in the efficiency of supply chains. These strategic problems involve stakeholders determining the optimal locations for warehouses to ensure effective distribution o...
详细信息
Warehouse location problems play a significant role in the efficiency of supply chains. These strategic problems involve stakeholders determining the optimal locations for warehouses to ensure effective distribution of commodities. However, other stakeholders are also involved in the supply chain. Hence, combining warehouse location decisions with operational decisions, such as distribution, inventory management or production planning, among others, allows for an integrated approach to supply chain management. In this study, we are focusing on stakeholders at two different levels: those in charge of warehouse locations and the managers of each warehouse, who are responsible for inventory decisions. Usually, these stakeholders are not aligned since each of them has their own interests. It is worth emphasizing that an obvious hierarchy among stakeholders exists. The decisions of the warehouse locator (leader) restrict the decision space of each warehouse manager (follower). Moreover, the inventory-related decisions made by each follower have an impact on the total cost of the supply chain. This hierarchical relationship between leader and followers can be modeled as a bi-level programming problem with multiple independent followers. To solve this complex problem, two nested evolutionary algorithms are proposed. The first one follows a traditional evolutionary scheme, and the second one is innovative, since similar individuals are grouped into sub-populations and a migration process is included to enhance the convergence of the algorithm. The nested component of the proposed algorithms is obtained by optimally solving each follower's problem using the Lagrangian multipliers method. Computational experimentation is conducted over a set of benchmark instances to validate the effectiveness and efficiency of the proposed nested evolutionary algorithms. Additionally, a case study is presented, and a sensitivity analysis regarding the number of warehouses to be opened is con
The rapid increase in the quantity of available biologic data over the last decade, brought about by the introduction of massively parallel methods for gene expression measurements, has highlighted the need for more e...
详细信息
The rapid increase in the quantity of available biologic data over the last decade, brought about by the introduction of massively parallel methods for gene expression measurements, has highlighted the need for more efficient computational techniques for analysis. This paper reviews the use of evolutionary algorithms (EAs) in connection with classification based on gene expression data matrices. Brief introductions to data classification methods and EAs are given, followed by a survey of studies dealing with the application of evolutionary algorithms to various (cancer related) data sets. The general conclusion, based on the published results surveyed here, is that EAs may constitute an efficient method for optimal gene selection, and can also help in reducing the size (number of features used) of classifiers. In many cases, the classification accuracy obtained using EAs, often in conjunction with other methods, represents a significant improvement over results obtained without the use of EAs. However, long-term, independent clinical follow-up studies will be essential to validate prognostic markers identified by the use of EA-based methods.
evolutionary algorithms are powerful search techniques which have been used successfully in many different domains. Parallel evolutionary algorithm has become a research focus due to its easy implement and promise sub...
详细信息
evolutionary algorithms are powerful search techniques which have been used successfully in many different domains. Parallel evolutionary algorithm has become a research focus due to its easy implement and promise substantial gains in performance. In this paper a framework of tree-modelbased parallel evolutionary algorithm (T-PEA) is proposed. The presented method employs Bayesian Dirichlet metric to construct a tree model from a set of potential solutions, which is then used to model potential solutions and guide exploration in the search space. The correctness and rationality of the proposed method for learning tree models are analyzed and proved in the context of genetic and evolutionary. The method is important not only for T-PEA, but also for machine learning and data mining. The experimental results show that the proposed algorithm can efficiently and rapidly converge and obtain the optimal solution from all test functions.
作者：
Abreu, NunoMatos, AnibalINESC
TEC Campus FEUPRua Dr Roberto Frias 378 P-4200465 Oporto Portugal FEUP
DEEC P-4200465 Oporto Portugal
Autonomous underwater vehicles (AUVs) are increasingly being used to perform mine countermeasures (MCM) operations but its capabilities are limited by the efficiency of the planning process. Here we study the problem ...
详细信息
Autonomous underwater vehicles (AUVs) are increasingly being used to perform mine countermeasures (MCM) operations but its capabilities are limited by the efficiency of the planning process. Here we study the problem of multiobjective MCM mission planning with AUVs. The vehicle should cover the operating area while maximizing the probability of detecting the targets and minimizing the required energy and time to complete the mission. A multi-stage algorithm is proposed and evaluated. Our algorithm combines an evolutionary algorithm (EA) with a local search procedure, aiming at a more flexible and effective exploration and exploitation of the search space. An artificial neural network (ANN) model was also integrated in the evolutionary procedure to guide the search. The combination of different techniques creates another problem, related to the high amount of parameters that needs to be tuned. Thus, the effect of these parameters on the quality of the obtained Pareto Front was assessed. This allowed us to define an adaptive tuning procedure to control the parameters while the algorithm is executed. Our algorithm is compared against an implementation of a known EA as well as another mission planner and the results from the experiments show that the proposed strategy can efficiently identify a higher quality solution set.
Although real-coded evolutionary algorithms (EAs) have been applied to optimization problems for over thirty years, the convergence properties of these methods remain poorly understood. We discuss the use of discrete ...
详细信息
Although real-coded evolutionary algorithms (EAs) have been applied to optimization problems for over thirty years, the convergence properties of these methods remain poorly understood. We discuss the use of discrete random variables to perform search in real-valued EAs. Although most real-valued EAs perform mutation with continuous random variables, we argue that EAs using discrete random variables for mutation can be much easier to analyze. In particular, we present and analyze two simple EAs that make discrete choices of mutation steps.
Optical networks are key enablers of the modern communication services to handle the increasing bandwidth requests. Virtualization is a feasible technology to response to the users' demands. On the other hand, the...
详细信息
Optical networks are key enablers of the modern communication services to handle the increasing bandwidth requests. Virtualization is a feasible technology to response to the users' demands. On the other hand, the cloud computing, as a distinguished use case of the virtualized optical networks, has impacted the IT world. In virtualization of the resources, optimum mapping of the virtual optical networks on the physical infrastructure plays an important role. In this paper, a novel optimized scheme for mapping the virtual optical networks on the physical infrastructure is proposed. A new formulation for Routing andWavelength Assignment, RWA, problem is presented. A novel encoding method for optical networks is proposed based on categorizing the wavelengths into different groups according to data transmission rate. Then, the Genetic Algorithm, GA, and Binary Particle Swarm Optimization, BPSO, as two popular evolutionary algorithms are implemented to find the optimum map of the virtual optical networks on the physical infrastructure using the proposed encoding method. The optimization constraints and two heuristics, proposed to satisfy them, are detailed. Finally, the simulation results for a physical infrastructure and different virtual optical networks are presented. Results show that the GA outperforms the BPSO in terms of providing optimized solutions with less values of the defined cost function. But the run time required to find the optimum map of the virtual optical networks is more than the BPSO.
暂无评论