Predicting the global minimum structures of atomic clusters has important practical implications in physics and chemistry. This is because the global minimum structures of their potential function theoretically correspond to their ground state structures, which determine some important physical and chemical properties of clusters. However, this prediction task is a very challenging global optimization problem due to the fact that the number of local minima on the potential energy surface of clusters increases exponentially with the cluster size. In this study, we propose an unbiased global optimization approach, called the iterated dynamic lattice search algorithm, to search for the global minimum structure of atomic clusters. Based on the iterated local search framework, the proposed algorithm employs the well-known monotonic basin-hopping method to improve the initial structures of clusters, a surface-based perturbation operator to randomly change the positions of selected surface atoms or central atom, a dynamic lattice search method to optimize the positions of surface atoms, and the Metropolis acceptance rule to accept the optimized new solutions. The performance of the algorithm is evaluated on the 300 widely studied silver clusters and experimental results show that the proposed algorithm is highly efficient compared to the existing algorithms. In particular, the proposed algorithm improves the best-known structures for 47 clusters and matches the best-known structures for the remaining clusters. Additional experiments are performed to analyze the key components of the algorithm and the landscape of the potential energy surface of several representative clusters.

To boost the search performance of Artificial Bee Colony (ABC) algorithm for handling some complicated optimization problems, a dual subpopulation ABC based on individual gradation (DPGABC) is presented. In DPGABC, the whole population is segmented into two subpopulations with different gradations. Then, the subpopulations respectively utilize the strategies with different characteristics as the candidate strategies. So the individuals can exploit the benefits of various strategies to optimize the search performance. Meanwhile, the dual subpopulation mechanism can maintain good population diversity while achieving good convergence performance. In addition, a knowledge-driven parameter update mechanism is designed to improve the convergence performance. The CEC2014 test set is applied for relevant experiments to validate the performance of DPGABC. From the results, DPGABC performs well on most functions.

To improve the convergence performance of artificial bee colony (ABC) algorithm for tackling some complex optimisation issues, a new ABC with multi-strategy adaptation (MSABC) is presented. A multi-strategy adaptation mechanism is implemented to boost the search performance in MSABC. In this mechanism, an evolution rate index is proposed to adaptively select strategies with different characteristics at various evolutionary stages. Meanwhile, for balancing exploration and exploitation, a novel search strategy oriented by the elite individual is applied in this mechanism. On the CEC2014 test set, MSABC is compared to other existing algorithms to assess its performance. As the results demonstrate, MSABC can obtain good convergence performance.

Reference point-based environmental selection has achieved promising performance in multi-objective optimization problems. However, when solving the irregular multi-objective optimization problems, the performance of environmental selection is affected. This is because the irregular Pareto front is often degraded, disconnected, inverted, or with sharp tails, resulting in some reference points not located in appropriate region. This releases the selection pressure. Therefore, adjusting or generating some points is necessary to tackle this problem. However, how to identify the region of interest and how to generate new points in the appropriate region are the current problems to be solved. In this paper, a region-based reconstruction for reference points is proposed. For simplicity, the smallest region which consists of M reference points (M is the dimension of objective space) in the hyperplane of reference point is identified as the unit region. If the vertexes of the region all belong to active reference points, the region will be identified as region of interest and new reference points will be reconstructed in this region. In addition, the process is activated in the later stage of the algorithm operation, while the efficient of the search algorithm is weak. In order to find more valuable individuals in the neighborhood region of selected individuals, thereby, firefly algorithm is employed as search algorithm because of its search mechanism which has strong indicative features. Several experiments are designed to verify the performance of the proposed method. The experiment results show that the proposed method is effective.

Considering the problem that the intermittent fault signal of the electronic system is greatly affected by noise and has a lot of redundant information, which results in the limitation of the deep neural network model to evaluate the severity of the intermittent fault. A method for evaluating the severity of intermittent faults based on Variational Mode Decomposition-Gated Recurrent Units (VMD-GRU) is proposed. Firstly, all Intrinsic Mode Function (IMF) components are adaptively decomposed on intermittent fault signals through Variational Mode Decomposition (VMD). Then the sensitivity analysis of the IMF components is performed to select the sensitive components, and the differential enhanced energy operator is used to construct the severity sensitivity factor. Finally, the severity sensitivity factor is used to train the Gated Recurrent Units (GRU) recurrent neural network severity evaluation model. Through the evaluation of intermittent faults of different severity injected into the key circuits of electronic systems, the results show that this method has a strong ability to evaluate the severity of intermittent faults, and is more accurate and effective in evaluating the severity of intermittent faults.

图像作为信息的主要载体,人们可从图像中获取更多后期决策的依据.由于获取图像的条件不同,导致在进行多幅图像融合时采取的方法不尽相同.如何从同一场景的不同图像中获取更为丰富的信息成为图像融合的关键问题.首先,对各种类型的图像融合进行总结得出图像融合的具体概念.其次,概述图像融合的经典理论及方法,根据处理对象将融合方法分为基于变换域的方法和基于空间域的方法并对其进行对比分析.接着,分析表征融合图像质量的主要评价标准特点和选取依据.最后,从医学诊断、遥感观测、摄影技术和监测应用等领域综述图像融合技术取得的突破,并重点对比分析各领域所用方法的优势与不足,探索现有融合方法的改进策略以及图像融合应用领域的推广.

The traditional gravitational search algorithm (GSA) maintains good diversity of solutions but often demonstrates weak local search ability. To promote the local search ability of GSA, a new GSA based on chaotic local search (CLSGSA) is introduced in this paper. In its search operations, CLSGSA first executes the conventional search operations of the basic GSA to maintain the diversity of solutions. After that, CLSGSA executes a chaotic local search with the search experience from the current best solution to increase the local search capability. In the experiments, we utilise a suite of benchmark functions to verify the performance of CLSGSA. Moreover, we compare the proposed CLSGSA with several GSA variants. The comparisons validate the effectiveness of CLSGSA.

In order to accelerate the convergence speed of the traditional DE algorithms for some complex optimization problems, an adaptive bare-bones differential evolution based on cosine (CABDE) was proposed. In the proposed CABDE, a new adaptive mechanism for mutation strategy selection was presented. Moreover, a cosine adaptive factor was introduced to achieve the complementary advantages of the Gaussian mutation strategy and the DE/current-to-best/1 mutation strategy. The Gaussian mutation strategy has excellent global search ability, which is good for maintaining the population diversity; While the DE/current-to-best/1 mutation strategy exhibits good local search ability, which is helpful for accelerating the convergence speed. Therefore, the presented adaptive mechanism can maintain a balance between the exploration and exploitation. In addition, the information of the best individual in both Gaussian mutation strategy and DE/current-to-best/1 mutation strategy was utilized to guide the search directions. During the evolution process, the cosine adaptive factor was adjusted according to the increase of the iterations and then the suitable mutation strategies in the different evolutionary stages were adaptively chosen. As a result, the presented adaptive mechanism can maintain the population diversity as well as accelerate the convergence speed. To test the performance of CABDE, 18 test functions with different characteristics were used in the experiments. The effectiveness of the mutation strategies and the dynamic parameters were discussed. The experimental results showed that the mutation strategies and the dynamic parameters can improve the search performance. Moreover, CABDE was compared with several new variants of bare-bones algorithms, DE variants, PSO variants, and ABC variants. The comparisons indicated that CABDE can achieve better solutions and exhibit faster convergence speed.

针对传统方法在解决化工参数辨识问题中易陷入局部最优、导致求解精度不足的问题,提出了一种组合三角变异差分进化(CTMDE)算法,融入了组合三角高斯变异策略和DE/current-to-pbest/1变异策略.其中,组合三角高斯变异策略引入了组合权重来适应性利用较优个体、一般个体、当前个体的信息,维持种群多样性;而DE/current-to-pbest/1变异策略能够利用种群中的较优个体来指导搜索,对解空间的开采能力较强.两者结合使得算法在加快收敛速度的同时降低陷入局部最优的可能性.在12个基准测试函数上,将CTMDE算法与其他新近DE算法进行比较,并将CTMDE算法应用于甲醇转化为烃类物质的参数辨识问题.实验结果表明:CTMDE算法具有较好的寻优性能,且在化工参数辨识问题上具有较好的求解效果.

在动态集成差分进化算法中,动态学习机制往往过于复杂且增加计算开销.为此,本文以传统差分进化算法框架为基础,提出集成DE/rand/1/bin和DE/best/1/bin两个优势互补的变异策略并设计动态执行机制,力求简化动态学习机制,且又能在全局搜索和局部搜索中寻找到平衡.实验结果表明:本文提出动态集成两个变异策略的差分进化算法(differential evolution algorithm integrated bi-mutation strategy,DE-BMS)缩放因子F选择为0.9,交叉概率Ct选择为0.1,算法具有更好的鲁棒性;与其他差分进化算法的收敛速度、成功次数解质量分别进行比较,DE-BMS在优化多峰函数问题时表现更佳.

Node deployment is a fundamental technique in wireless sensor networks, which can be converted into an optimisation issue. Gravitational Search Algorithm (GSA) is a popular optimisation method, which has exhibited promising performance for node deployment. However, the traditional GSA may show poor convergence when tackling some complicated node deployment issues. To enhance the search efficiency, an enhanced GSA (NDGSA) is introduced for node deployment issue in wireless sensor networks. In NDGSA, it creates new solutions according to a linear combination of the current solution and the value drawn from Gaussian distribution. In the experiments, NDGSA is compared with the traditional approaches on the node deployment issue. The comparisons validate the efficiency of NDGSA.

Artificial bee colony (ABC) algorithm is a promising evolutionary algorithm inspired by the foraging behaviour of honey bee swarm, which has obtained satisfactory solutions in diverse applications. However, the basic ABC often demonstrates insufficient exploitation capability in some cases. To address this concerning issue, a chaotic artificial bee colony with elite opposition-based learning strategy (CEOABC) is proposed in this paper. During the search process, CEOABC employs the chaotic local search to promote the exploitation ability. Moreover, the elite opposition-based learning strategy is utilised to exploit the potential information of the exhausted solution. Experimental results compared with several ABC variants show that CEOABC is a competitive approach for global optimisation.

Harmony search (HS) has been widely used in the field of wireless sensor networks. However, the search strategy of the basic HS has excellent exploration capability but weak exploitation capability. To enhance the search capability of HS, this paper presents a hybrid harmony search algorithm (HBHS) for node localisation in wireless sensor networks. The proposed HBHS employs the best solution to enhance the exploitation capability. Moreover, HBHS utilises an adaptive search step-size scheme to further enhance the search capability. To verify the search performance, HBHS is compared with two HS algorithms on a suit of classical benchmark problems. The comparisons confirm that HBHS can achieve better performance than the compared HS algorithms on the most of the benchmark problems. Further, HBHS is applied for node localisation in wireless sensor networks.

In evolutionary multi-objective optimization, achieving a balance between convergence speed and population diversity remains a challenging topic especially for many-objective optimization problems (MaOPs). To accelerate convergence toward the Pareto front and maintain a high degree of diversity for MaOPs, we propose a new many-objective dynamical evolutionary algorithm based on E-dominance and adaptive-grid strategies (EDAGEA). In EDAGEA, it incorporates the E_dominance and adaptive strategies to enhance the search ability. Instead of the Pareto dominance mechanism in the traditional dynamical evolutionary algorithm, EDAGEA employs the E-dominance strategy to improve the selective pressure and to accelerate the convergence speed. Moreover, EDAGEA incorporates the adaptive-grid strategy to promote the uniformity and diversity of the population. In the experiments, the proposed EDAGEA algorithm is tested on DTLZ series problems under 3–8 objectives with diverse characteristics and is compared with two excellent many-objective evolutionary algorithms. Experimental results demonstrate that the proposed EDAGEA algorithm exhibits competitive performance in terms of both convergence speed and diversity of population.

As a powerful evolutionary algorithm for solving the tough global optimization problems, differential evolution (DE) has drawn more and more attention. However, how to make a proper balance between the global and local search is a perplexing question and greatly limit the optimization performance of DE. As we all known, there are two classical mutation strategies in DE, i.e., DE/rand/1 and DE/best/1. In DE/rand/1 strategy, the base vector is chosen from the population randomly, this means its better exploration and poor exploitation. The base vector of DE/best/1 strategy is the best one of the population and the strategy has better exploitation and poor exploration. To overcome these problems, this paper proposed a random neighbor based mutation strategy (DE/neighbor/1). For each individual of the population at each generation, the neighbors are chosen from the population in a random manner. The base vector of DE/neighbor/1 mutation strategy is the best one in the neighbors. On the basis of the new strategy, an enhancing differential evolution with DE/neighbor/1 (RNDE) is proposed. The experimental studies have been tested on 27 widely used benchmark functions and the results have proved that the proposed algorithm is competitive and promising.

Differential evolution (DE) is a simple yet efficient stochastic search approach for numerical optimization. However, it tends to suffer from slow convergence when tackling complicated problems. In addition, its search ability is significantly influenced by its control parameters. To improve the performance of the basic DE, this paper proposes a self-adaptive differential evolution with global neighborhood search (NSSDE). In the proposed NSSDE, its control parameters are self-adaptively tuned according to the feedback from the search process, while the global neighborhood search strategy is incorporated to accelerate the convergence speed. To evaluate the performance of the proposed NSSDE, we compare it with several DE variants on a set of benchmark test functions. The experimental results show that NSSDE can achieve better results than its competitors on the majority of the benchmark test functions.

The evolution strategy (ES) based on covariance matrix adaptation (CMA) is an excellent,gradient-free stochastic local optimization method.The learning mechanism based on CMA enables evolution strategy algorithm to have invariance to any invertible linear transformation of the search space,and to have outstanding capability for solving the illconditioned and/or highly non-separable problems.The learning mechanism of CMA has a solid theoretical foundation in mathematics,which may have a certain reference significance to guide the design of other evolutionary algorithms.This paper aims at analyzing the learning mechanisms of CMA-ES in detail,and providing its main mathematical foundations.Finally,the advantages and disadvantages of various CMA-ES variants are compared by a series of experiments,and the difference in performance is compared seriously between our improved variant and other CMA-ES variants.

The bank account location (BAL) problem is an NP-hard discrete optimization problem. A few experimental studies have shown that evolutionary algorithms are efficient methods for the BAL problem. However, from theoretical point of view, we know little about the performance of evolutionary algorithms (EAs) on the BAL problem. In this paper, we contribute to theoretical understanding of EAs on the BAL problem. The worst-case bounds on a simple evolutionary algorithm called (1 + 1) EA and a global simple multiobjective evolutionary algorithm called GSEMO for the BAL problem is presented. We reveal that the (1 + 1) EA can find a (k/(2k - 1)) approximation solution for the BAL problem. We also find that GSEMO can obtain an approximate solution on the BAL problem with value not less than (1-(1/e))OPT in expected polynomial runtime O(n2 log n + nk2), where OPT is the optimal fitness function value, n is the number of banks that can open accounts, and k is the maximum number of accounts that can be maintained. Meanwhile, we demonstrate that the (1+1) EA and GSEMO are superior to some local search algorithms with interchange neighborhood on an instance, and we also show that GSEMO can efficiently optimize another instance while the (1 + 1) EA may be inefficient.

Gravitational search algorithm (GSA) is an emerging evolutionary algorithm (EA), which has exhibited remarkable performance in many applications. However, the traditional gravitational search algorithm tends to yield slow convergence speed when facing some complicated real-life problems. Aiming at this weakness, a new gravitational search algorithm with Gaussian mutation strategy (GMGSA) is presented. At each generation, GMGSA calculates the centre of the current individual and the global best individual, and then combines the obtained centre information into the Gaussian mutation strategy to generate new individuals. In the experiments, GMGSA is evaluated on a set of well-known benchmark problems. The experimental results indicate that GMGSA can demonstrate promising performance.

Gravitational search algorithm (GSA), a popular evolutionary computation technique, has been widely employed in data management. However, the basic GSA demonstrates good exploration search but weak exploitation search. To promote the exploitation capability of GSA, this paper introduces a modified GSA with crossover (CROGSA). In its search process, CROGSA executes the crossover-based search scheme to update the position of each solution. Moreover, the crossover-based search scheme takes advantage of the promising knowledge extracted from the global optimal position achieved until now to enhance the exploitation capability. Experiments on a suit of benchmark cases indicate that CROGSA is better than several related optimization approaches in most of the cases. (C) 2017 Elsevier Ltd. All rights reserved.