查看更多>>摘要：The continuous roaming of particles in high-dimensional space makes it difficult for particle swarm optimization to achieve better optimization results. On the other hand, increasing the dimensionality may also bring about an explosive increase in the number of locally optimal solutions surrounded by more significant local optimal regions. Therefore, the algorithm requires high convergence while maintaining good diversity. This paper proposes a level-based multi-strategy learning swarm algorithm called LSLSO. LSLSO's optimization process is divided into the level-based multi-strategies search and the detailed search stages. First, particles are divided into four levels according to their fitness value. When the particles are at different levels, the particles have different learning strategies to update. Particles with better fitness focus on exploiting space. In contrast, particles with poor fitness will focus on exploring space. In the detailed search stage, particles at the same level learn from each other. Particles with similar fitness to detail search for the small gaps in the promising space already explored. The theoretical discussion results show that LSLSO has strong competitiveness in exploration and exploitation capabilities. Moreover, it shows good performance compared with the most advanced large-scale multi-objective optimization algorithms on the LSMOP and LMF problems.
查看更多>>摘要：In contemporary software development, developers commonly conduct regression testing to ensure that code changes do not affect software quality. Conducting performance regression testing after every code change is known to be expensive which emigres the need to direct the performance regression testing efforts on code changes that are most likely introducing performance regression. In this paper, we exploit code change metrics to identify the group introducing the regression based on a pre-trained detection rule. We present PRICE approach as a new formulation of detecting code changes introducing performance regression as an optimization problem using multi-objective evolutionary algorithms. PRICE evaluated using a set of 8000 commits, extracted from the Git project. Results show the effectiveness of our approach in accurately detecting performance regression introducing code changes. The average regression detection PRICE provides is 77% which is 22% more than the state-of-the-art deterministic approach. This improvement didn't compromise the detection of the contradicted code changes that are not introducing a regression. PRICE provides detection of code changes not introducing a regression averaged 63%, which is also an improvement of 14% than the comparative approach. Using PRICE, we were able to explore new search spaces and provide competing results.
查看更多>>摘要：The core task of constrained multiobjective optimization is to achieve a tradeoff between exploration and exploitation as well as a tradeoff between constraints and objectives. We present an effective evolutionary algorithm with a dynamic population size (DPSEA) to achieve these two tradeoffs. In order to balance exploration and exploitation, the population size of DPSEA decreases continually as generation increases. In this manner, a bigger population size can encourage exploration in the early stage, while a smaller one can promote exploitation in the later stage. Aiming to balance constraints and objectives, a two-stage environmental selection strategy is proposed. In this strategy, objective information and constraint information is used to select two sets of solutions, respectively. Note that the sizes of these two sets are also adjusted dynamically. In this way, the information of both objectives and constraints can be used. Moreover, a novel mating selection strategy is designed to select promising parents. By assembling the above processes, DPSEA is able to achieve the two tradeoffs which are critical to constrained multiobjective optimization. Experiments on three sets of benchmark test functions owning difficult characteristics validate that DPSEA is competitive against some state-of-the-art constrained multiobjective optimization evolutionary algorithms.
查看更多>>摘要：During the last decade, metaheuristic algorithms have been well-established approaches which are utilized for solving complex real-world optimization problems. The most metaheuristic algorithms uses stochastic strategies in their initialization phase as well as during their new candidate generation steps when there is no a-priori knowledge about the solution, which is a common valid assumption for any black-box optimization algorithm. In recent years, researchers have introduced a new concept called center-based sampling which can be utilized in any search component of the optimization process, but so far it mainly has been used just for population initialization. This novel concept clarifies that in a search space, the center point has a higher chance to be closer to an unknown solution compared to a random point, especially when the dimension of the search space increases. Thus, this concept helps the optimizer to find a better solution in a shorter time. In this paper, a comprehensive study has been conducted on the effect of center-based sampling to solve an optimization problem using three different levels of detailed investigation. These levels are as the follows: 1) considering no specific algorithm and no specific landscape (i.e., Monte-Carlo simulation); 2) considering a specific landscape but no specific algorithm (i.e., random search vs. center-based random search), and finally, 3) considering a specific algorithm and specific landscape which includes the proposing three different schemes for using center-based sampling scheme for solving Large-scale Global Optimization (LSGO) problems effectively. Furthermore, in this study, we seek to investigate the properties and capabilities of center-based sampling during optimization, which can be extended to utilize it in machine learning too, because optimization is a key role player in search and learning models. The proposed methods in this paper are evaluated on CEC 2013 LSGO benchmark functions and a real-world optimization problem, i.e., evolving ANN on two medical data sets. The experimental results confirm that center -based sampling has a crucial impact in improving the convergence rate of optimization/search algorithms when solving high-dimensional optimization problems.
查看更多>>摘要：This paper presents a grammar-based evolutionary approach that facilitates autonomous discovery of abstrac-tions to learn complex collective behaviours through manageable sub-models. We propose modifications to the design of the genome structure of the evolutionary model and the grammar syntax to facilitate representation of abstractions in separate partitions of a genome. Two learning architectures based on parallel and incremen-tal learning are proposed to automatically derive abstractions. The evaluations conducted with three different complex task environments indicate that the proposed approach with both architectures surpass the performance of generic grammar-based evolutionary models by automatically identifying appropriate abstractions and gen-erating more complex rule structures. The evolutionary process shows further performance improvements with the use of scaffolded environments which were used to train the models in increasingly complex environments across several stages. The results infer that the proposed approach incorporating grammatical evolution with techniques to autonomously discover abstractions can facilitate solving complex problems of agent systems in real-world domains.
查看更多>>摘要：In this paper, a two-stage surrogate-assisted evolutionary algorithm (TS-SAEA) is presented for computationally expensive multi/many-objective optimization, which consists of a convergence stage and a diversity stage. In the convergence stage, the objective space is partitioned into several sub-regions by reference vectors, where the individuals compete with each other. In the diversity stage, the converged individuals and the current non dominated solutions are combined to form a potential sample set, on which a secondary selection is conducted to further improve the diversity. Specifically, the proposed diversity strategy firstly defines the initial boundary individuals and a candidate pool. The individuals with "max-min angles " will continuously be selected from the pool to supplement the boundary individuals until the number of the boundary individuals equals the number of the current non-dominated solutions. At last, the points with the better space-filling features are picked out from the updated boundary individuals to evaluate the true objectives. The above-mentioned process keeps running until the maximal number of function evaluations is satisfied. To evaluate the performance of TS-SAEA on both low and high-dimensional multi/many-objective problems, it is compared with four state-of-art algorithms on 52 benchmark problems and one engineering application. The experimental results show that TS-SAEA has significant advantages on computationally expensive multi/many-objective optimization problems.
查看更多>>摘要：Multiobjective evolutionary algorithms (MOEAs) have faced the challenge of balancing diversity and convergence in dealing with many-objective optimization problems (MaOPs). Most of them use a series of strategies to increase the selection pressure among solutions for convergence promotion, or additional auxiliary strategies for diversity maintenance. Decision variable classification (DVC), as the method that analyzes the feature of an MaOP, can help MOEAs search for optimal solutions in terms of convergence and diversity through optimizing the corresponding category of decision variables. Therefore in this paper, we propose a new DVC method by analyzing the monotonicities of objectives. Unlike other DVC methods, it does not need to consider dominance relationships or help with extra vectors. Based on the classification results, we design a new directional crossover (DC) method for generating promising solutions. This crossover method has a higher probability that the generated offspring can integrate the advantages of the parents in convergence and diversity. Incorporating it with MOEA, a DVC-based MOEA (DVC-MOEA) is proposed for dealing with MaOPs. In DVC-MOEA, two archives focusing on convergence and diversity separately are maintained. In addition, an interval mapping(IM) strategy is designed to obtain solutions with good diversity, especially for some problems with biased features. To evaluate the performance of DVC-MOEA on MaOPs, comparison experiments are conducted on two wide used benchmarks with nine state-of-the-art MOEAs. The experimental results show that DVC-MOEA has high competitiveness over these MOEAs in dealing with MaOPs. Moreover, three variants are compared with DVC-MOEA respectively, and the comparison experimental results confirm the effect of the three strategies (DVC, DC, and IM) in our proposed algorithm.
查看更多>>摘要：During an escaped wildfire in a populated area's vicinity, protective tasks should be carried out to secure crucial community assets, e.g., bridges, hospitals, power stations, and communication towers. In a real-life scenario, an important asset may require the combined effort of different fire suppression resources, which should be dispatched and scheduled to act synchronously in protecting the respective asset. The present research addresses the solution of a challenging routing problem in emergency response, the Asset Protection Problem (APP), which incorporates selective characteristics in routing a heterogeneous vehicle fleet with complex temporal and spatial constraints, i.e., time windows and synchronization requirements. Notably, the Modified Ant Colony System (MACS) algorithm is proposed to obtain effective APP solutions within a time suitable for operational purposes. Based on the conducted experiments, MACS outperforms the previously published solution approaches in the solution of large-scale APP benchmark instances. Notably, MACS obtained superior solutions in 159 out of 240 large-scale instances, while 87 of them represent new best results, considering the solutions achieved by the commercial solver CPLEX with a ten-hour time limit.
查看更多>>摘要：Large-scale sparse multi-objective optimization problems exist widely in the real world, but most existing evolutionary algorithms encounter great difficulties in solving the problems of this type, mainly due to the curse of dimensionality and the underutilized sparsity knowledge of the Pareto optimal solutions. To address these issues, this paper proposes a multi-stage knowledge-guided evolutionary algorithm for large-scale sparse multi-objective optimization problems, which aims to enhance the optimization capability by incorporating diversified sparsity knowledge into the evolutionary process. Specifically, three kinds of the knowledge are designed and an effective multi-stage evolutionary strategy based on knowledge fusion is developed to make full use of three kinds of knowledge. Experimental results on eight benchmark problems and three real-world problems demonstrate that the proposed algorithm outperforms the state-of-the-art approaches in terms of effectiveness and convergence speed.
查看更多>>摘要：The truss optimization problem has been extensively investigated, and the optimized trusses have been widely used in various fields. Truss optimization is a challenging optimization involving many difficulties, such as discrete variables and non-convex problems. In order to solve these problems, various metaheuristic optimization methods have been proposed. Due to the stochastic feature of these methods, it is always computationally intensive, and the optimization results may vary greatly in different optimization runs. In this paper, a hybrid intelligent genetic algorithm (HIGA) is proposed to improve the effectiveness and efficiency of truss optimization problems. This method systematically integrates deep neural network (DNN) and genetic algorithm (GA) in the optimization process. A two-step training procedure is proposed where the data generated during the optimization process is exploited to update DNN. Next, based on the generated DNN model, an optimization prediction procedure is proposed to seek more optimized trusses in an efficient manner. Through the investigation of three classical truss problems (size optimization, shape optimization and size-shape integrated optimization), the effectiveness of the proposed method is validated. In addition, the influence of different settings on the optimization performance and efficiency is investigated to demonstrate the applicability and robustness of the proposed method.
NSTL
Elsevier