Constrained multi-objective optimization problems often exhibit complex and fragmented feasible regions, which makes global constrained Pareto front (CPF) search challenging. To address this, we present an unbalanced and balanced competitive strategy-assisted dual-swarm optimizer (UBCSO). UBCSO introduces a balanced swarm that uses random pairing and even grouping to accelerate convergence toward feasible regions, and an unbalanced, cluster-based swarm divides each cluster into three fixed groups for intra-cluster learning, promoting diverse exploration. Within clusters, we propose an unbalanced competition strategy to achieve more precise particle selection during learning, and a cluster-based environmental selection enables effective information exchange across clusters and swarms, enhancing CPF search. Additionally, the proposed cluster integration strategy enables efficient use of computing resources by reducing exploration in non-promising regions. Finally, across 37 instances, UBCSO attained the best overall ranking compared to ten popular methods (IGD+ 1.7838, HV 1.984) and achieved the best IGD+ metric in 25 out of 37 problems.

Constrained multi-objective evolutionary algorithms (CMOEAs) typically integrate diverse evolutionary operators, constraint-handling techniques, and environmental selection (ES) strategies to address constrained multi-objective optimization problems (CMOPs). Notably, significant performance variations emerge when identical CMOEAs employ different operators for solving the same CMOP, a phenomenon arising from distinct operator preferences exhibited by CMOPs with varied landscape. Therefore, it is worthwhile and promising to adaptively select appropriate operators for different CMOPs. This paper proposes CMOFLCS, a novel framework incorporating a fuzzy logic control system (FLCS). We introduce dual metrics for assessing population convergence and diversity, combined with a reward mechanism that dynamically evaluates operators' historical contributions. These metrics feed into the FLCS, which synergizes expert-defined rules with real-time data feedback to probabilistically select optimal operators via roulette wheel. Furthermore, we develop an angle-constrained ES that redirects inefficient exploration of the unconstrained Pareto front (UPF) to a uniform search of the objective space in the UPFto- constrained Pareto front (CPF) direction. This mechanism activates adaptively when handling problems with complete UPF-CPF separation. Experiments on 33 benchmark problems and 25 real-world applications demonstrate that CMOFLCS achieves superior performance, outperforming eight state-of-theart CMOEAs. Significantly, FLCS integration enhances baseline performance when embedded into two popular CMOEAs, further validating both its effectiveness and generalizability.

Distributed hybrid flow shop scheduling is prevalent in industries such as integrated circuit manufacturing, ceramic frit production, glass fiber processing, and steelmaking. Machine breakdowns and deteriorating jobs represent common and disruptive sources of uncertainty in these distributed manufacturing environments. However, existing research has often overlooked these significant challenges. To address this gap, this article addresses the dynamic distributed hybrid flow shop scheduling problem with machine breakdowns and deteriorating jobs (DHFSP-MBDJs), and develops the mathematical model. We propose a learning-assisted genetic programming hyperheuristic (L-GP-HH) algorithm to minimize makespan. L-GP-HH incorporates a novel constructive heuristic for factory assignment and develops specific terminal sets based on fundamental factors and uncertain events to generate genetic programming (GP) heuristics. Additionally, we establish a learning probability model to optimize the selection of GP-generated rules during the solution process. Extensive numerical experiments demonstrate that L-GP-HH consistently outperforms conventional GP hyperheuristics (GP-HHs), benchmark scheduling rules, and four efficient meta-and hyperheuristics. It exhibits superior flexibility and adaptability in handling complex scheduling under dynamic environment with uncertainties. This study provides critical insights for practitioners, emphasizing the necessity of concurrently considering machine-and job-related uncertainties in dynamic distributed manufacturing systems to enhance scheduling robustness and operational efficiency.

As an important field leading the rapid development of China’s economy, industry is an important support for building a modern power, and it is also a large carbon emitter in China. Therefore, it is of key significance to promote industry to achieve the peak of carbon emissions for the realization of China’s “dual-carbon goals”. Aiming at the problem of distributed heterogeneous flow shop scheduling problem based on dual-carbon goals (DHFSP-DCGs), a novel distributed heterogeneous flow shop scheduling model was constructed to minimize the maximum completion time and total carbon emissions, and a knowledge-driven multi-objective memetic algorithm was proposed. Firstly, considering the machine characteristics of heterogeneous factories and the conflict between two optimization objectives, the encoding and decoding methods based on double sequences are designed. Secondly, a cooperative initialization strategy is proposed to generate the initial solutions with good diversity and convergence. Thirdly, according to the characteristics of distributed heterogeneous flow shop scheduling problem, a knowledge-based local search strategy is designed to improve the quality of the solution and the performance of the algorithm, and carbon reduction strategy is used to reduce the carbon emission in the production scheduling process. Finally, the effectiveness of the proposed strategy and algorithm is verified by comparative experimentsNote to Practitioners—This paper is to solve the problem of distributed heterogeneous flow shop scheduling problem based on dual-carbon goals (DHFSP-DCGs), and the methods proposed could bring many benefits to practitioners. Firstly, considering the machine characteristics of heterogeneous factories and the conflict between two optimization objectives, the encoding and decoding methods based on double sequences are designed. Secondly, a cooperative initialization strategy is proposed to generate the initial solutions with good diversity and convergence. Thirdly, according to the characteristics of distributed heterogeneous flow shop scheduling problem, a knowledge-based local search strategy is designed to improve the quality of the solution and the performance of the algorithm, and carbon reduction strategy is used to reduce the carbon emission in the production scheduling process.

Multimodal remote sensing image segmentation (MRSIS) is important for intelligent remote sensing image (RS) interpretation, which encompasses three distinct tasks: semantic segmentation, instance segmentation, and panoptic segmentation. Existing methods typically address individual tasks with specialized models, limiting generalization and real-world applicability. Multi-task learning approaches have introduced separated task heads to unify tasks, yet we identify two key challenges when directly applying them to MRSIS: (1) the modality gap, arising from semantic discrepancies and granularity discrepancies across RS modalities, and (2) the task gap, due to varying preferences in learning different segmentation tasks. To overcome these challenges, we propose PMTSeg—a novel Prompt-driven Multimodal Transformer for task-adapted MRSIS. PMTSeg integrates three key components: (1) Task-common Multimodal Affinity Approximation (TMAA), (2) Task-common Multi-scale Semantic Fusion (TMSF), and (3) a unified Prompt-driven Segmentation Head (PSH). First, TMAA addresses the modality gap by approximating inter-modal affinity matrices, extracting task-common features across modalities and aligning semantic information. Then, TMSF further integrates these features using the scale-matched fusion at multiple scales to produce enriched, multi-scale task-common features. Moreover, to address the task gap, the PSH leverages task-adapted text prompts and task-adapted contrastive loss to model relationships across tasks, enabling adaptive optimization for robust and universal MRSIS performance. Extensive experiments on three MRSIS datasets—VALID, SEMCITY TOULOUSE, and UBCV2—demonstrate that PMTSeg significantly surpasses state-of-the-art methods in all three segmentation tasks, offering a unified and accurate solution to MRSIS.

Cloud computing has emerged as a dominant force in tackling dynamic, unpredictable, and adaptable computing demands. However, the vast scale of cloud computing, the complexities of diverse scenarios, and the unpredictable nature of user requests pose significant challenges to achieving efficient and effective cloud computing scheduling. To address these issues and enhance computational efficiency while reducing energy consumption in cloud computing systems, we introduce a workload-adaptive deep reinforcement learning algorithm that dynamically adjusts the discount factor in accordance with real-time workload changes. To demonstrate the effectiveness of our proposed method, we utilize a real dataset from the Google Cloud computing center, which encompasses workloads from 12,000 machines running over 670,000 applications and 40 million jobs. Simulation results reveal that our method achieves a 9.93% reduction in task rejection rate and a 5.01% decrease in energy consumption cost compared to state-of-the-art solutions.

Using seismic information for oil and gas exploration can improve the accuracy of exploration, which makes this technology become the research focus of oil and gas exploration. The inversion technology based on prestack seismic wave, which contains more elastic parameter information reflecting the characteristics of underground reservoirs, has become a popular technology in seismic inversion. In view of the shortcomings of traditional deep learning methods in the inversion of elastic parameters of pre-stack seismic waves, this paper proposes an improved combination method of convolutional neural network, gated recurrent unit and attention mechanism, which can comprehensively and deeply mine the data and obtain better prediction effect through the fusion of multiple models. The experimental results show that the proposed method can obtain the solution with higher inversion precision and has a good application prospect in the complex seismic wave impedance.

Summary Graph data processing has been widely applied in a variety of domains such as industry, science, social network, and so on. It therefore has stimulated many efforts devoted to this area. To embrace the fast development trend of big graph data, graph data processing based on Pregel‐like systems has been regarded as one of the most promising ways and has widely attracted the attention of researchers. However, it still remains in its early stage and there still exist many challenges. In Pregel, the superstep synchronization is time consuming as the graph data iteration operation requires multiple synchronizations. Furthermore, the graph data partition strategy adopted by Pregel fails to support load balancing, therefore causing the increase of network I/O overhead as the scale of graph data grows. To address these issues, this paper presents an efficient computational framework for graph data processing based on the bulk synchronous parallel model. The global synchronization control mechanism is improved by determining the start time of the next round of superstep through counting the number of global message files. Furthermore, an improved graph data partition mechanism based on a balanced hash method is proposed to reduce the communication overhead between different partitions of sub‐graph computational tasks. We also re‐design the PageRank algorithm to verify the effectiveness of the proposed framework. Experimental results on different real‐world datasets verify the efficiency of our proposed framework as it outperforms Giraph (an open source Pregel‐like system) by 58%−69%, and achieves 10×−17× performance improvement over Hadoop.

Summary Burst or leakage in drinkable water distribution system has occurred frequently in recent years, causing severe damages, economic loss, and long‐lasting society impact. A viable solution is to use agile inline mobile sensors to detect and so as to mitigate the burst or leakage. Distinguishing from online fixed sensors, mobile sensors can swim freely along the piles in water distribution network, thus giving a more precise detection. To combat the low power, low computation, and low communication capability of mobile sensors, the newly emerged fog computing provides a promising means to gather and preprocess the sensing data. In practice, due to the budget limitation, we can deploy a limited number of sensors and fog nodes in the system. This introduces a challenging problem on how to deploy them in the system, ie, sensor and fog node placement. We first formulate mobile sensor placement (MSP) as a path cover problem and prove it as NP‐complete, and then we propose a customized genetic algorithm and a mixed greedy algorithm to solve MSP and fog node placement, respectively. The correctness and efficiency of the proposed algorithm are illustrated by a comprehensive experiment. Moreover, some critical factors, eg, sensor battery lifetime and movement pattern, are all extensively investigated and the results show the coverage ratio is sensitive to these factors.

Nonlinear equations system (NES) arises commonly in science and engineering. Repulsion techniques are considered to be the effective methods to locate different roots of NES. In general, the repulsive radius needs to be given by the user before the run. However, its optimal parameter setting is difficult and problem-dependent. To alleviate this drawback, in this paper, we first propose a dynamic repulsion technique, and then a general framework based on the dynamic repulsion technique and evolutionary algorithms (EAs) is presented to effectively solve NES. The major advantages of our framework are: 1) the repulsive radius is controlled dynamically during the evolutionary process; 2) multiple roots of NES can be simultaneously located in a single run; 3) the diversity of the population is preserved due to the population reinitialization; and 4) different repulsion techniques and different EAs can be readily integrated into this framework. To extensively evaluate the performance of our framework, we choose 42 problems with diverse features as the test suite. In addition, some representative differential evolution and particle swarm optimization variants are incorporated into the framework. Our method is also compared with other state-of-the-art methods. Experimental results indicate that the dynamic repulsion technique can improve the performance of the original repulsion technique with static repulsive radius. Moreover, the proposed method is able to yield better results compared with other methods.

Summary The elastic parameter inversion technique for prestack seismic data, which combines the intelligent optimization algorithms with Amplitude Variation with Offset (AVO) technology, is an effective method for oil and gas exploration. However, when certain biological‐evolution–based optimization algorithms, eg, genetic algorithms, are used to solve this problem, the computation exhibits fast convergence and a strong tendency to be trapped to a local optimum, thereby leading to unsatisfactory inversion results. To address this issue, this paper proposes a swarm‐intelligence‐based method‐Particle Swarm Optimization (PSO) algorithm to handle the elastic parameter inversion problem. Based on the Aki‐Richards approximation to the Zoeppritz equations, the improved PSO algorithm adopts a special initialization strategy, which can enhance the smoothness of the initialization parametric curves. Extensive experimental research confirms the superiority of the proposed algorithm. Specifically, the improved PSO algorithm is able to not only markedly enhance inversion precision but also render remarkably high correlation coefficients associated with the elastic parameters.

Hyperspectral remote sensing images present rich information on the characteristics of different physical materials. Utilizing the rich information, classifiers can distinguish these different materials. The minimum distance technique, which is commonly used in classification, is sensitive to the distance metric, especially in high-dimensional space. In this letter, we study the effect of the $p$ -norm distance metric on the minimum distance technique and propose a supervised-learning $p$ -norm distance metric to optimize the value of $p$ . In the experimental study, we take the minimum distance and the $k$ -nearest neighbor classifiers as examples to test the proposed supervised-learning $p$ -norm distance metric. The results suggest that the supervised-learning $p$ -norm distance metric can improve the performance of a classifier for hyperspectral remote sensing image classification.

Summary In recent years, the drinking water pollution incident occurred frequently, a serious threat to social stability and security. By using the sensor networks real‐time monitoring the urban water supply networks, the water pollution event probability can be greatly reduced. But knowing how to use the water quality monitor sensor networks to collect information to identify pollution sources is still a challenging problem. In this paper, we formulate the contaminant source identification problem into an optimization problem, and then design the cultural algorithm to solve it by considering different sizes of water supply networks as the experimental data. Finally, the experimental results verify the effectiveness and robustness of the proposed method.

Multi-label classification is to assign an instance to multiple classes. Naive Bayes (NB) is one of the most popular algorithms for pattern recognition and classification. It has a high performance in single label classification. It is naturally extended for multi-label classification under the assumption of label independence. As we know, NB is based on a simple but unrealistic assumption that attributes are conditionally independent given the class. Therefore, a double weighted NB (DWNB) is proposed to demonstrate the influences of predicting different labels based on different attributes. Our DWNB utilizes the niching cultural algorithm (NLA) to determine the weight configuration automatically. Our experimental results show that our proposed DWNB outperforms NB and its extensions significantly in multi-label classification.

The personalized and diverse demands of modern communication present new challenges to antenna design. While the emergence of Software-Defined Everything provides an innovative hardware design idea that hardware structure is modeled in a software way and designed with intelligence optimization algorithms. Inspired by the design idea, in this paper we propose a software-defined intelligent method for antenna design. The optimal variables are described with the software-defined antenna structure method, and the optimal target is determined with the software-defined antenna performance method. Based on the above the software-defined antenna model will be abstracted, which convert the work of antenna design to the optimal problem on antenna structure. Meanwhile considering the features of the long simulation and the intensive computation, an intelligent algorithm is proposed suitable for the intelligent design method. This software-defined intelligent method is applied to an antenna design example and several optimal antennas satisfying the requirements have been obtained. It demonstrates that automation and intelligentization in antenna design can be achieved with this innovative method.

Software defined wireless sensor network can be adapted to different application needs through dynamic programming. In this paper, we propose a signal acquisition and processing wireless sensor network (SAPSN). SAPSN consists of sampling nodes, processing nodes and remote controllers. At first, the sampling node completes the local signal sampling by analog-digital conversion. Next, according to the different demand from the remote controller, processing node completes time domain or frequency domain analysis of signal processing, and transfers the results back to the remote controller. Finally, the application in remote controller will display the results according to different user's needs. SAPSN is capable of time domain or frequency domain signal analysis and processing, depending on different application requirements. In this paper, we present the concept underlying SAPSN, its architecture. We also present preliminary experimental results.

In practical applications, the interference between the antenna and the metal structure nearby often cause deviation of the electromagnetic performance of the antenna. The factor of finite reflector should be considered in the evolutionary antenna to avoid the influence so that the optimal antenna structure can be better applicable to real-world conditions. Therefore, a kind of design model of evolutionary antenna with a finite reflector is proposed in this paper, which includes the corresponding chromosome coding and the design workflow. In this design model, the antenna is optimized with its finite reflector during the evolutionary process. Meanwhile aiming at the characteristics of the model, the method of invoking the simulation software and setting the finite reflector is discussed to evaluate the antenna individuals effectively. This design model with a finite reflector was applied in a practical design problem, and the experimental result shows that it is superior to the one with a infinite reflector both on gain and VSWR. It illustrates that it is very important to take the reflector factor into consideration in the design model of evolutionary antenna.

Multi-label classification is a generalization of single-label classification, and its samples belong to multiple labels. The K-nearest neighbor algorithm can solve this problem as an optimization problem. It finds the optimum solution by caculating the distance between each sample in general. But in fact, the distance of K-nearest neighbor algorithm may be miscalculated due to the caused by the redundant or irrelevant characteristic value. In order to solve this problem, in this paper, we propose a novel method that uses the particle swarm algorithm to optimize the feature weights to improve the accuracy of distance calculation. As a result, it can improve classification accuracy further. The experimental results show that applying particle swarm algorithm's optimization technique to improving K-nearest neighbor algorithm for multi-label classification problem, can improve the accuracy of classification effectively.