Paradigm Of Evolutionary Algorithms Research Articles

An artificial bee colony algorithm with a cumulative covariance matrix mechanism and its application in parameter optimization for hearing loss detection models

Artificial bee colony (ABC) algorithms which are applied to various complex problems for their competitive performance and simple structures are a relatively popular paradigm of evolutionary algorithms (EAs). Although many variants based on the basic ABC algorithm have been developed, most variants have limited performance on non-separable problems. This is mainly because these variants usually focus on only individual quality information, but neglect population distribution information. To address this limitation, in this paper, we propose a new ABC with a cumulative covariance matrix mechanism (CCoM-ABC), in which the population distribution information in all previous generations is involved to guide the evolution process. Specifically, the CCoM-ABC algorithm applies a cumulative covariance matrix (CCoM) mechanism by which the population information of every generation is involved to improve the probability of finding the optimum. It also applies an adaptive selection mechanism by which the proper coordinate system used to generate candidate solutions can be dynamically selected between a natural one and an eigen one during the search process. Furthermore, for improving the search ability, a new search equation is adopted in the scout bees phase. A series of experiments are conducted on the CEC2014 to evaluate the performance of the proposed CCoM-ABC algorithm, and the results show that the CCoM-ABC outperforms state-of-the-art ABC variants and EAs in terms of convergence speed, solution accuracy, and numerical scalability. For validating the practicality of the proposed CCoM-ABC algorithm, we further apply it to optimize the parameters of the neural network for hearing loss (HL) detection. This study involves 60 brain images consisting of 20 left-sided HL, 20 right-side HL, and 20 healthy control images. The results show that this model outperforms state-of-the-art related methods in terms of overall accuracy by about 2%.

ERG-DE: An elites regeneration framework for differential evolution

Differential evolution (DE) is one of the most popular paradigms of evolutionary algorithms. Numerous variants of basic DE have been developed in the past two decades after it was first proposed. However, very few works focused on re-exploring the neighborhood area of the elite solutions, which is definitely a promising area according to the proximate optimality principle. Here, a simple yet efficient elites regeneration (ERG) framework was designed to fill this gap. The elite population in this framework is defined as a group of individuals with better fitness values and they are regenerated after the selection procedure in DE. Specifically, a new individual is produced from the search space around each elite individual (i.e. the parent individual) by sampling Gaussian or Cauchy probability models and replaces the parent if it has better fitness value. The implementation of this procedure only introduces two parameters that need to be tuned, i.e. the standard deviation for Gaussian distribution and the scale parameter for the Cauchy distribution. In the proposed framework, the elite individuals serve as the mean or location parameters of the probability models and the standard deviation and scale parameters are tuned by experiments as a small constant value. Thus, offspring individuals are generated in areas close to their corresponding elite parents. The framework allows thorough exploitation of search neighborhoods around elite individuals and ultimately helps the elite individuals escaping from local optima. Experiments results on CEC2014 benchmark revealed that ERG framework significantly increased the optimization capacity for four original DE algorithms, four classical DE variants, and two state-of-the-art DE variants. In addition, it also demonstrated competitive performance when compared with another DE framework.

Exploring self-organization of molecular tether molecules on a gold surface by global structure optimization.

We employ nondeterministic global cluster structure optimization, based on the evolutionary algorithms paradigm, to model the self-assembly of complex molecules on a surface. As a real-life application example directly related to many recent experiments, we use this approach for the assembly of triazatriangulene "platform" molecules on the Au(111) surface. Without additional restrictions like spatial discretizations, coarse-graining or precalculated adsorption poses, and despite the proof-of-principle character of this study, we achieve satisfactory qualitative agreement with several experimental observations and can provide answers to questions that experiments on these species had left open so far. © 2019 Wiley Periodicals, Inc.

A Two-Stage Ensemble of Differential Evolution Variants for Numerical Optimization

Differential evolution (DE) is a popular paradigm of evolutionary algorithms, which has been widely applied to solve diverse optimization problems and has gained much success in a series of academic benchmark competitions. Recently, ensemble methods received an increasing attention in designing high-quality DE algorithm. Motivated by this consideration, a novel two-stage ensemble of DE variants, called TSEDE, has been proposed in this paper. In TSEDE, based on the number of fitness evaluations, the whole evolutionary process is divided into two stages. In the former stage, TSEDE using a multi-population-based framework focuses on improving the searchability, which employs three popular and efficient DE variants, namely SHADE, JADE, and “DE/current-to-rand/1.” In the latter stage, LSHADE is used to emphasize the convergence. Moreover, an elite strategy is used to ensure that the current best solution is assigned to each constituent variant at each generation. TSEDE is tested on the CEC2005 benchmark suit and compared with nine typical algorithms. The results confirm that the proposed method is very competitive.

Utilizing cumulative population distribution information in differential evolution

Differential evolution (DE) is one of the most popular paradigms of evolutionary algorithms. In general, DE does not exploit distribution information provided by the population and, as a result, its search performance is limited. In this paper, cumulative population distribution information of DE has been utilized to establish an Eigen coordinate system by making use of covariance matrix adaptation. The crossover operator of DE implemented in the Eigen coordinate system has the capability to identify the features of the fitness landscape. Furthermore, we propose a cumulative population distribution information based DE framework called CPI-DE. In CPI-DE, for each target vector, two trial vectors are generated based on both the original coordinate system and the Eigen coordinate system. Then, the target vector is compared with these two trial vectors and the best one will survive into the next generation. CPI-DE has been applied to two classic versions of DE and three state-of-the-art variants of DE for solving two sets of benchmark test functions, namely, 28 test functions with 30 and 50 dimensions at the 2013 IEEE Congress on Evolutionary Computation, and 30 test functions with 30 and 50 dimensions at the 2014 IEEE Congress on Evolutionary Computation. The experimental results suggest that CPI-DE is an effective framework to enhance the performance of DE.

Prior knowledge guided differential evolution

Differential evolution (DE) has become one of the most popular paradigms of evolutionary algorithms. Over the past two decades, some prior knowledge has been obtained in the DE research community. It is an interesting topic to enhance the performance of DE by taking advantage of such prior knowledge. Along this line, a prior knowledge guided DE (called PKDE) is proposed in this paper. In PKDE, we extract two levels of prior knowledge, i.e., the macro- and micro-levels. In order to integrate these two levels of prior knowledge in an effective way, the control parameters of PKDE are tuned based on two distributions (i.e., Cauchy and normal distribution), with the aim of alleviating the premature convergence at the early stage and speeding up the convergence toward the global optimal solution at the later stage. In addition, the self-adaptive mutation strategy is implemented based on our previous study. PKDE is compared with eight DE variants and seven non-DE algorithms on two sets of benchmark test functions from IEEE CEC2005 and IEEE CEC2014. Systematic experiments demonstrate that the overall performance of PKDE is very competitive.

Incorporating PLS model information into particle swarm optimization for descriptor selection in QSAR/QSPR

As a representative paradigm of evolutionary algorithms, particle swarm optimization (PSO) has been combined with partial least square (PLS) (called PSO‐PLS) to select informative descriptors in quantitative structure‐activity/property relationship (QSAR/QSPR). However, one of the main limitations of PSO‐PLS is that it ignores PLS model information. In this paper, by incorporating the PLS model information into PSO‐PLS, we present a novel weighted sampling method (called WS‐PSO‐PLS) to choose the optimal descriptor subset. Due to the fact that the regression coefficients of the PLS model reflect the importance of descriptors in the model development, we firstly obtain the normalized regression coefficients by establishing the PLS model with all the descriptors. Afterward, weighted sampling is used to generate some individuals according to the aforementioned normalized regression coefficients. Finally, we employ some dimensions of the generated individuals to replace the corresponding dimensions of the individuals with poor quality in the population at each generation. WS‐PSO‐PLS has been assessed through three QSAR/QSPR datasets and the experimental results suggest that WS‐PSO‐PLS has the capability to effectively guide the search process by introducing the PLS model coefficients into PSO during the evolution and, therefore, performs better than PSO‐PLS. WS‐PSO‐PLS could be considered as a general and promising mechanism to introduce extra information to improve the performance of PSO for descriptor selection in QSAR/QSPR. Copyright © 2015 John Wiley & Sons, Ltd.

Semi-supervised evolutionary ensembles for Web video categorization

Evolutionary Algorithms (EA) have been developing rapidly as a powerful and general learning approach which has been used successfully to find a reasonable solution for data mining and knowledge discovery. Genetic algorithm (GA) is a kind of mainstream EA paradigm with a purpose of developing solutions for optimization problems. Clustering ensembles have emerged as an outstanding algorithm in machine learning to leverage the consensus across multiple clustering solutions and combines their predictions into a single solution with improved robustness, stability and accuracy. Multimedia advancement and popularity of the social Web has collectively provided an easy way to generate bulk of videos. Categorization of such Web videos has become a hot research challenge. In this paper, we propose a Semi-supervised Evolutionary Ensemble (SS-EE) framework for social media mining, e.g., Web Video Categorization (WVC), using their low cost textual features, intrinsic relations and extrinsic Web support. The contributions of this research work are as follows. First, we extend the traditional Vector Space Model (VSM) to Semantic VSM (S-VSM) by considering the semantic similarity between the feature terms using Normalized Google Distance (NGD) approach. Second, we define a new distance measure, Triangular Similarity (TrS) between two Textual Feature Vectors (TFV) based on the frequencies of most relevant terms in each category. Third, we iterate the clustering ensemble process with the help of GA guided by a new measure, Pre-Paired Percentage (PPP), to be used as the fitness function during the genetic cycle. Fourth, in the key steps of the GA, crossover and mutation genetic operators, we define them by an intelligent mechanism of clustering ensemble. Fifth, in order to terminate the genetic cycle, we define another new measure, Clustering Quality (Cq), based on similarity matrix and clustering labels. Experiments on real world social-Web data (YouTube) have been performed to validate the SS-EE framework.

Neutrality in evolutionary algorithms… What do we know?

Over the last years, the effects of neutrality have attracted the attention of many researchers in the Evolutionary Algorithms (EAs) community. A mutation from one gene to another is considered as neutral if this modification does not affect the phenotype. This article provides a general overview on the work carried out on neutrality in EAs. Using as a framework the origin of neutrality and its study in different paradigms of EAs (e.g., Genetic Algorithms, Genetic Programming), we discuss the most significant works and findings on this topic. This work points towards open issues, which we belive the community needs to address.

Evolutionary model trees for handling continuous classes in machine learning

Model trees are a particular case of decision trees employed to solve regression problems. They have the advantage of presenting an interpretable output, helping the end-user to get more confidence in the prediction and providing the basis for the end-user to have new insight about the data, confirming or rejecting hypotheses previously formed. Moreover, model trees present an acceptable level of predictive performance in comparison to most techniques used for solving regression problems. Since generating the optimal model tree is an NP-Complete problem, traditional model tree induction algorithms make use of a greedy top-down divide-and-conquer strategy, which may not converge to the global optimal solution. In this paper, we propose a novel algorithm based on the use of the evolutionary algorithms paradigm as an alternate heuristic to generate model trees in order to improve the convergence to globally near-optimal solutions. We call our new approach evolutionary model tree induction (E-Motion). We test its predictive performance using public UCI data sets, and we compare the results to traditional greedy regression/model trees induction algorithms, as well as to other evolutionary approaches. Results show that our method presents a good trade-off between predictive performance and model comprehensibility, which may be crucial in many machine learning applications.

Adaptive Reservoir Evolutionary Algorithm: An Evolutionary On-Line Adaptation Scheme for Global Function Optimization

This paper introduces a novel global optimization heuristic algorithm based on the basic paradigms of Evolutionary Algorithms (EA). The algorithm greatly extends a previous strategy proposed by the authors in Munteanu and Lazarescu (1998). In the newly designed algorithm the exploration/exploitation of the search space is adapted on-line based on the current features of the landscape that is being searched. The on-line adaptation mechanism involves a decision process as to whether more exploitation or exploration is needed depending on the current progress of the algorithm and on the current estimated potential of discovering better solutions. The convergence with probability 1 in finite time and discrete space is analyzed, as well as an extensive comparison with other evolutionary optimization heuristics is performed on a set of test functions.

Two aspects of evolutionary algorithms

In this paper we discuss the paradigm of evolutionary algorithms (EAs). We argue about the need for new heuristics in real-world problem solving, discussing reasons why some problems are difficult to solve. After introducing the main concepts of evolutionary algorithms, we concentrate on two issues: (1) self-adaptation of the parameters of EA, and (2) handling constraints.