Swarm and Evolutionary Computation最新文献

Dynamic and multimodal properties are simultaneously possessed in the dynamic multimodal optimization problems (DMMOPs), which aim to find multiple optimal solutions in a dynamic environment. However, more work still needs to be devoted to solving DMMOPs, which still require significant attention. A niching-based brain storm optimization with two archives (NBSO2A) algorithm is proposed to solve DMMOPs. The two niching methods, i.e., neighborhood-based speciation (NS), and nearest-better clustering (NBC), are incorporated into a BSO algorithm to generate new solutions. The two archives preserve the optimal solutions that meet the requirements and practical, inferior solutions discarded during the generation. Improved taboo area (ITA) removes highly similar individuals from the population. An evolution strategy with covariance matrix adaptation (CMA-ES) is adopted to enhance the local search ability and improve the quality of the solutions. The NBSO2A algorithm and four other algorithms were tested on 12 benchmark problems to validate the performance of the NBSO2A algorithm on DMMOPs. The experimental results show that the NBSO2A algorithm outperforms the other compared algorithms on most tested benchmark problems.

Dynamic multiobjective optimization problems (DMOPs) widely appear in various real-world applications and have attracted increasing attention worldwide. However, how to obtain both good population diversity and fast convergence speed to efficiently solve DMOPs are two challenging issues. Inspired by that the multiple populations for multiple objectives (MPMO) framework can provide algorithms with good population diversity and fast convergence speed, this paper proposes a new efficient algorithm called a co-evolutionary multi-population evolutionary algorithm (CMEA) based on the MPMO framework together with three novel strategies, which are helpful for solving DMOPs efficiently from two aspects. First, in the evolution control aspect, a convergence-based population evolution strategy is proposed to select the suitable population for executing the evolution in different generations, so as to accelerate the convergence speed of the algorithm. Second, in the dynamic control aspect, a multi-population-based dynamic detection strategy and a multi-population-based dynamic response strategy are proposed to help the algorithm maintain the population diversity, which are efficient for detecting and responding to the dynamic changes of environments. Integrating with the above strategies, the CMEA is proposed to solve the DMOP efficiently. The superiority of the proposed CMEA is validated in experiments on widely-used DMOP benchmark problems.

Many real world optimization problems are formulated as mixed-variable optimization problems (MVOPs) which involve both continuous and discrete variables. MVOPs including dimensional variables are characterized by a variable-size search space. Depending on the values of dimensional variables, the number and type of the variables of the problem can vary dynamically. MVOPs and variable-size MVOPs (VMVOPs) are difficult to solve and raise a number of scientific challenges in the design of metaheuristics. Standard metaheuristics have been first designed to address continuous or discrete optimization problems, and are not able to tackle VMVOPs in an efficient way. The development of metaheuristics for solving such problems has attracted the attention of many researchers and is increasingly popular. However, to our knowledge there is no well established taxonomy or comprehensive survey for handling this important family of optimization problems. This paper presents an unified taxonomy for metaheuristic solutions for solving VMVOPs in an attempt to provide a common terminology and classification mechanisms. It provides a general mathematical formulation and concepts of VMVOPs, and identifies the various solving methodologies than can be applied in metaheuristics. The advantages, the weaknesses and the limitations of the presented methodologies are discussed. The proposed taxonomy also allows to identify some open research issues which needs further in-depth investigations.

In decomposition-based multi-objective evolutionary algorithms (MOEAs), the inconsistency between a problem’s Pareto front shape and the distribution of the weights can lead to a poor, unevenly distributed solution set. A straightforward way to overcome this undesirable issue is to adapt the weights during the evolutionary process. However, existing methods, which typically adapt many weights at a time, may hinder the convergence of the population since changing weights essentially means changing sub-problems to be optimised. In this paper, we aim to tackle this issue by designing a steady-state weight adaptation (SSWA) method. SSWA employs a stable approach to maintain/update an archive (which stores high-quality solutions during the search). Based on the archive, at each generation, SSWA selects one solution from it to generate only one new weight while simultaneously removing an existing weight. We compare SSWA with eight state-of-the-art weight adaptative decomposition-based MOEAs and show its general outperformance on problems with various Pareto front shapes.

Discrete & continuous optimization constitutes a challenging task and generally rises as an NP-hard problem. In the literature, as a derivative of this type of optimization issue, hyperparameter optimization of transfer learning (TL) architectures is not efficiently analyzed as a detailed survey in the literature. In this paper, the optimized TL-based models are effectively examined to handle this issue which constitutes the main aim of our study. For evaluation, knee osteoarthritis (KOA – a chronic degenerative joint disorder) dataset is handled to perform two challenging classification tasks which reveal the second aim of our study, i.e. binary- and multi-categorizations on KOA X-ray images. To fine-tune the hyperparameters of TL models, state-of-the-art optimization methods are chosen and compared on this competitive – NP-hard problem. Sixteen optimized architectures are designed using four efficient optimization methods (ASPSO, CDW-PSO, CSA, MSGO) and four oft-used TL models (MobileNetV2, ResNet18, ResNet50, ShuffleNet) to classify the X-ray KOA images. Regarding the experiments on both categorization tasks, the MSGO algorithm arises as more robust to be considered for hyperparameter tuning of TL-based models by achieving reliable performance. In addition, it's seen that MobileNetV2 and ResNet-based models come to the forefront for X-ray imaging-based classification by achieving high accuracy rates due to the usage of residual blocks. Consequently, in terms of mean accuracy, ResNet50-MSGO and MobileNetV2-CSA respectively record 93.15 % and 93.29 % success rates on multiclass categorization, while ResNet18-CDW-PSO and MobileNetV2-MSGO provide the same highest score of 99.43 % on binary categorization.

Production scheduling and maintenance planning are two interactive factors in modern manufacturing system. However, at present, almost all studies ignore the impact of unpunctual maintenance activities on the integrated production and maintenance scheduling since the unavailabilities of repairmen are dynamically changed, e.g., repairmen increase, decrease and their unavailable intervals update. Under these contexts, this paper addresses a novel integrated optimization problem of condition-based preventive maintenance (CBPM) and production rescheduling with multi-phase processing speed selection and dynamic repairman assignment. More precisely, (1) a novel multi-phase-multi-threshold CBPM policy with remaining-useful-life-based inspection and multi-phase processing speed selection is proposed to obtain some selectable maintenance plans for each production machine; (2) a hybrid rescheduling strategy (HRS) that includes three rescheduling strategies is designed for responding to the dynamic changes of repairman; and (3) an adaptive clustering- and Meta-Lamarckian learning-based bi-population co-evolutionary algorithm (ACML-BCEA) is developed to deal with the concerned problem. In the numerical simulations, the effectiveness of designed operators and proposed ACML-BCEA algorithm is first verified. Next, the superiority and competitiveness of the proposed CBPM policy and HRS are separately demonstrated by comparing with other CBPM policies and rescheduling strategies. After that, a comprehensive sensitivity analysis is performed to analyze the effect of optional range of processing speed, skill level of selectable repairmen and total number of processing phases, and the analyzing results show that these factors all have a significant impact on the integrated optimization.

Most studies in dynamic multi-objective optimization have predominantly focused on rapidly and accurately tracking changes in the Pareto optimal front (POF) and Pareto optimal set (POS) when the environment undergoes changes. However, there are real-world scenarios where it is necessary to simultaneously solve changing objective functions and satisfy the preference of Decision Makers (DMs). In particular, the DMs may be only interested in a partial region of the POF, known as the region of interest (ROI), rather than requiring the entire POF. To meet the challenge of simultaneously predicting a changing POF and/or POS and dynamic ROI, this paper proposes a new dynamic multi-objective evolutionary algorithm (DMOEAs) based on the preference. The proposed algorithm consists of three key components: an evolutionary direction adjustment strategy based on changing reference points to accommodate shifts in preferences, an angle-based search strategy for tracking the varying ROI, and a hybrid prediction strategy that combines linear prediction models and population manifold estimation within the ROI to ensure convergence and distribution in scenarios where preferences remain unchanged. Experimental studies conducted on 30 widely used benchmark problems in which it outperforms contrasting algorithms on 71% of test suits. Empirical results demonstrate the significant advantages of the proposed algorithm over existing state-of-the-art DMOEAs.

Dynamic constrained multi-objective optimization problems (DCMOPs) involve objectives, constraints, and parameters that change over time. This kind of problem presents a greater challenge for evolutionary algorithms because it requires the population to quickly track the changing pareto-optimal set (PS) under constrained conditions while maintaining the feasibility and good distribution of the population. To address these challenges, this paper proposes a dynamic constrained multi-objective optimization algorithm based on co-evolution and diversity enhancement (CEDE), in which we have made improvements to both the static optimization and dynamic response parts, innovatively utilizing the valuable information latent in the optimization process to help the population evolve more comprehensively. The static optimization involves the co-evolution of three populations, through which their mutual synergy can more comprehensively identify potential true PS and provide more useful historical information for dynamic response. Additionally, to prevent the elimination of potentially valuable infeasible individuals (i.e., individuals that are not dominated by feasible individuals) due to pareto domination, we employ an archive set to store and update these individuals. When the environment changes, to effectively enhance population diversity under complex dynamic constraints and help the population to respond quickly to changes, we propose a diversity enhancement strategy, which includes a diversity maintenance strategy and a center point-based exploration strategy. This strategy effectively enhances population diversity in complex and changing environments, helping the population respond quickly to changes. The effectiveness of the algorithm is validated through two test sets. The experimental results show that CEDE can effectively use valuable information to cope with complex dynamic constraint environments. Compared with several of the most advanced algorithms, it is superior in 94% of the test problems, demonstrating strong competitiveness in handling DCMOPs.

To improve the effectiveness of surrogate-assisted evolutionary algorithms (SAEAs) in solving high-dimensional expensive optimization problems with multi-polar and multi-variable coupling properties, a new approach called DRBM-ASRL is proposed. This approach leverages restricted Boltzmann machines (RBMs) for feature learning and reinforcement learning for adaptive strategy selection. DRBM-ASRL integrates four search strategies based on three heterogeneous surrogate modeling approaches, each catering to different preferences. Two of these strategies focus on generative sampling in the subspaces with varying dimensions, while the other two aim to explore the local and global landscapes in the high-dimensional source space. This allows for more effective tradeoffs between exploration and exploitation in the solution space. Reinforcement learning is employed to adaptively prioritize the search strategies during optimization , based on the online feedback information from the optimal solution. In addition, to enhance the representation of potentially optimal samples in the solution space, two task-driven RBMs are separately trained to construct a feature subspace and reconstruct the features of the source space. DRBM-ASRL has been evaluated on various high-dimensional benchmarks ranging from 50 to 200 dimensions, as well as 14 CEC 2013 complex benchmark problems with 100 dimensions and a power system problem with 118 dimensions. Experimental results demonstrate its superior convergence performance and optimization efficiency compared to eight state-of-the-art SAEAs.