IEEE Transactions on Emerging Topics in Computational Intelligence最新文献_第6页

UMIC: Super-Resolution of Cine Cardiac MRI Using U-Shaped Network With Multi-Level Information Compensation UMIC：基于多级信息补偿的u型网络的超分辨率电影心脏MRI

IF 6 3区计算机科学

IEEE Transactions on Emerging Topics in Computational Intelligence Pub Date : 2026-04-01 Epub Date: 2025-11-20 DOI: 10.1109/TETCI.2025.3631689

Defu Qiu;Kelvin K.L. Wong;Yuhu Cheng;Zhang Yi;Xuesong Wang

{"title":"UMIC: Super-Resolution of Cine Cardiac MRI Using U-Shaped Network With Multi-Level Information Compensation","authors":"Defu Qiu;Kelvin K.L. Wong;Yuhu Cheng;Zhang Yi;Xuesong Wang","doi":"10.1109/TETCI.2025.3631689","DOIUrl":"https://doi.org/10.1109/TETCI.2025.3631689","url":null,"abstract":"In cine cardiac magnetic resonance imaging (CMRI), deep learning-based super-resolution (SR) reconstruction algorithms often suffer from feature information loss during feature extraction and lack effective mechanisms for feature compensation. These problems can lead to the lack of texture and edge details in the reconstructed image, making it difficult to obtain a clear cardiac image, which will increase the rate of misjudgment of cardiac disease by experts. To address these issues, we propose a U-shaped network with multi-level information compensation (UMIC). Specifically, the network first performs multi-level feature extraction on low-resolution (LR) inputs and reduces channel dimensionality via a downward channel branch. The compressed features are then fused through a bottom module to capture inter-channel dependencies. Finally, the relevant features are recovered and enhanced through an upward channel branch. Additionally, we introduce a multi-level information compensation module to mitigate detail loss incurred during channel compression and to assist in recovering difficult-to-restore LR image details in the reconstruction phase. Extensive experiments show that UMIC achieves better CMRI SR reconstruction performance compared to some state-of-the-art SR methods.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"10 2","pages":"1987-1998"},"PeriodicalIF":6.0,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148511944","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A Spatiotemporal Graph Neural Network Based Deep Reinforcement Learning Method for Online Distributed Job-Shop Scheduling Problem 基于时空图神经网络的在线分布式作业车间调度深度强化学习方法

IF 6 3区计算机科学

IEEE Transactions on Emerging Topics in Computational Intelligence Pub Date : 2026-04-01 Epub Date: 2025-11-24 DOI: 10.1109/TETCI.2025.3628751

Jiangping Huang;Liang Gao;Xinyu Li

{"title":"A Spatiotemporal Graph Neural Network Based Deep Reinforcement Learning Method for Online Distributed Job-Shop Scheduling Problem","authors":"Jiangping Huang;Liang Gao;Xinyu Li","doi":"10.1109/TETCI.2025.3628751","DOIUrl":"https://doi.org/10.1109/TETCI.2025.3628751","url":null,"abstract":"The distributed shop scheduling problem is a hotspot in the shop scheduling field. Online scheduling requires making prompt decisions in response to environmental changes during ongoing production. This paper studies an Online Distributed Job-shop Scheduling Problem (ODJSP) with random job arrivals and machine breakdowns. The spatiotemporal Graph Neural Network (GNN) and Deep Reinforcement Learning (DRL) method are integrated, introducing a novel framework for efficiently solving online scheduling problems. First, a dynamic disjunctive graph with an adaptable topology is introduced. Building upon this foundation, a Markov Decision Process model is developed to formulate the ODJSP. A novel time-window-based state observation mechanism is designed, enhancing the agent's decision efficiency by curtailing redundant input information. Additionally, the Graph Convolutional Recurrent Network is employed to extract features during decision process, enabling effective feature capture across both spatial and temporal domains. The features across the temporal domain are crucial for online scheduling problems but have rarely been considered in existing research. Moreover, Proximal Policy Optimization is integrated with an actor-critic framework to train the decision agent. Comparative experiments are conducted across 972 simulation environments with different configurations. The comparison among 29 composite Priority Dispatch Rules, 3 GNN-based and Multi-Layer Perceptron-based DRL methods, and Deep Q-learning Network-based and Gene Expression Programming-based online scheduling methods demonstrates the effectiveness, stability, generalization, and real-time performance of the proposed method. The case study from a marine opto-mechanical structural component manufacturing company validates its practical value.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"10 2","pages":"1201-1215"},"PeriodicalIF":6.0,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148512144","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Microscale-Searching Evolutionary Optimization for Image Matting 图像抠图的微尺度搜索进化优化

IF 6 3区计算机科学

IEEE Transactions on Emerging Topics in Computational Intelligence Pub Date : 2026-04-01 Epub Date: 2025-12-24 DOI: 10.1109/TETCI.2025.3637791

Li Kang;Han Huang;Yihui Liang;Zeyang Liu;Zhifeng Hao

{"title":"Microscale-Searching Evolutionary Optimization for Image Matting","authors":"Li Kang;Han Huang;Yihui Liang;Zeyang Liu;Zhifeng Hao","doi":"10.1109/TETCI.2025.3637791","DOIUrl":"https://doi.org/10.1109/TETCI.2025.3637791","url":null,"abstract":"Image matting is a fundamental computer vision task. It is a large-scale optimization problem involving the selection of pixel pairs from pixel-pair sets of an image. Less prior, such as the trimaps of images, is required for evolutionary algorithms (EAs) to solve the image matting problem compared to deep learning-based methods. However, it is challenging for EAs to solve the image matting problem efficiently due to the large size of the decision set. This paper proposes a framework to guide EAs to search in a microscale subset of the decision set. The subset is estimated by collecting best-so-far solutions during EAs solve similar subproblems. Experimental results demonstrated that by integrating the proposed framework, EAs require fewer FEs to achieve competitive results compared to their original versions. Additionally, the results also indicate that the proposed strategy enhance the performance of EAs in terms of mean squared error and connectivity metrics compared to other EAs-based methods in most cases. The contribution of our work is to make EAs efficient algorithms to solving the image matting problem in scenarios with weak prior.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"10 2","pages":"1726-1737"},"PeriodicalIF":6.0,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148513689","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Knowledge Transfer-Based Co-Evolutionary Search for Energy-Saving Distributed Heterogeneous Flexible Job Shop Scheduling 基于知识转移的协同进化搜索节能分布式异构柔性作业车间调度

IF 6 3区计算机科学

IEEE Transactions on Emerging Topics in Computational Intelligence Pub Date : 2026-04-01 Epub Date: 2025-09-18 DOI: 10.1109/TETCI.2025.3607390

Tao Li;Xingchen Li;Junna Zhang;Dong Liu;Zidong Wang

{"title":"Knowledge Transfer-Based Co-Evolutionary Search for Energy-Saving Distributed Heterogeneous Flexible Job Shop Scheduling","authors":"Tao Li;Xingchen Li;Junna Zhang;Dong Liu;Zidong Wang","doi":"10.1109/TETCI.2025.3607390","DOIUrl":"https://doi.org/10.1109/TETCI.2025.3607390","url":null,"abstract":"This paper investigates the energy-saving distributed heterogeneous flexible job shop scheduling problem (ES-DHFJSP), with the dual objectives of reducing both makespan and total energy consumption. To address ES-DHFJSP, a knowledge transfer-based co-evolutionary search algorithm (KTCE) is proposed. Three key components are incorporated into the algorithm. First, a greedy population initialization strategy is designed, by which the characteristics of multi-factory distributed systems are leveraged to generate superior initial individuals. Second, a dynamic neighborhood search structure centered on critical factory is developed, through which the algorithm's capability to explore promising regions around elite solutions is enhanced. Third, a knowledge transfer mechanism centered on a factory allocation probability model is implemented, wherein the optimal factory allocations from previous generations are utilized to construct the probability model and the evolution of subsequent population is guided. The algorithm's convergence rate is substantially improved by this mechanism. The effectiveness and efficiency of KTCE are validated through comprehensive experimental results on various benchmark instances, where superior performance is demonstrated in solving ES-DHFJSP.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"10 2","pages":"1610-1624"},"PeriodicalIF":6.0,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148508184","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Large-Scale Constrained Multiobjective Optimization Based on Variable Adaptive Optimization and Population Reconstruction 基于变量自适应优化和种群重建的大规模约束多目标优化

IF 6 3区计算机科学

IEEE Transactions on Emerging Topics in Computational Intelligence Pub Date : 2026-04-01 Epub Date: 2025-12-17 DOI: 10.1109/TETCI.2025.3641668

Kunjie Yu;Zhenyu Yang;Jing Liang;Kangjia Qiao;Boyang Qu;Qiao Peng;Ponnuthurai Nagaratnam Suganthan

{"title":"Large-Scale Constrained Multiobjective Optimization Based on Variable Adaptive Optimization and Population Reconstruction","authors":"Kunjie Yu;Zhenyu Yang;Jing Liang;Kangjia Qiao;Boyang Qu;Qiao Peng;Ponnuthurai Nagaratnam Suganthan","doi":"10.1109/TETCI.2025.3641668","DOIUrl":"https://doi.org/10.1109/TETCI.2025.3641668","url":null,"abstract":"Constrained multiobjective evolutionary algorithms (CMOEAs) have been proposed to address constrained multiobjective optimization problems (CMOPs), and they have shown promising performance on small-scale problems. However, for CMOPs with large-scale decision variables, complex coupling relationships among variables and the huge search space make the algorithm converge slowly and difficult to find feasible solutions. Therefore, this study proposes a large-scale CMOEA based on variable adaptive optimization and population reconstruction to better solve large-scale CMOPs. First, a variable adaptive optimization strategy is developed in the global search stage, where convergence-related and diversity-related variables are adaptively selected for optimization. This strategy reduces the variable optimization difficulty caused by complex coupling relationships and explores the entire search space to locate regions near the constrained Pareto front (CPF). Then, in the local search stage, an archive-assisted population reconstruction strategy is proposed to generate an initial population by using promising solutions from the global search stage to rapidly guide the population towards the CPF. The experiments are conducted on 43 benchmark problems with up to 1000 decision variables and five real-world problems. Results show that the proposed method has better performance than other latest algorithms.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"10 2","pages":"1625-1640"},"PeriodicalIF":6.0,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148508186","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Fast and Slow Diffusion With Fuzzy Contextual Representation for Medical Image Translation 基于模糊上下文表示的快速和慢扩散医学图像翻译

IF 6 3区计算机科学

IEEE Transactions on Emerging Topics in Computational Intelligence Pub Date : 2026-04-01 Epub Date: 2025-12-18 DOI: 10.1109/TETCI.2025.3642000

Yu Tang;Jiahao Zheng;Fangmin Ren;Dan Huang;Dan Zeng;Dapeng Wu

{"title":"Fast and Slow Diffusion With Fuzzy Contextual Representation for Medical Image Translation","authors":"Yu Tang;Jiahao Zheng;Fangmin Ren;Dan Huang;Dan Zeng;Dapeng Wu","doi":"10.1109/TETCI.2025.3642000","DOIUrl":"https://doi.org/10.1109/TETCI.2025.3642000","url":null,"abstract":"The goal of medical image translation tasks is to design models that transform source domain images to target domain images, with the potential to significantly enhance diagnostic accuracy and reduce patient radiation exposure. Current advanced methods employ generative models to characterize the conditional distribution of target domain images, typically incorporating high frequency edge information from source domain images as generation conditions to maintain anatomical consistency between the two domains. However, directly introducing edge information may introduce noise and misleading information, adversely affecting the generative model. To address this issue, we propose a Fast and Slow Diffusion framework with Fuzzy Contextual Representation. Specifically, High-frequency edge images undergo fast diffusion while original images follow a slow diffusion process, coupled by a contextual pool that dynamically stores edge patterns from the fast path. Furthermore, we apply fuzzification to the contextual pool to enhance its expressiveness and use the resulting fuzzy representations to guide the reverse process of the slow diffusion, ensuring the accurate preservation of edge structures. Comprehensive experiments demonstrate that our proposed framework achieves performance comparable to existing state-of-the-art methods in medical image translation tasks. In addition, quantitative experiments demonstrate that fuzzified edge representations enhance generative models' capability in precisely perceiving and retaining high-frequency components.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"10 2","pages":"1469-1481"},"PeriodicalIF":6.0,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148508270","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Face Sketch Synthesis via Mamba-Transformer Encoder and Style-Enhanced Wavelet Convolution 基于Mamba-Transformer编码器和风格增强小波卷积的人脸草图合成

IF 6 3区计算机科学

IEEE Transactions on Emerging Topics in Computational Intelligence Pub Date : 2026-04-01 Epub Date: 2025-12-23 DOI: 10.1109/TETCI.2025.3641985

Mingzhang Liu;Weiguo Wan;Yuansheng Luo;Yong Gu;Jianjian Yan;Li Yao;Xinwu Li

{"title":"Face Sketch Synthesis via Mamba-Transformer Encoder and Style-Enhanced Wavelet Convolution","authors":"Mingzhang Liu;Weiguo Wan;Yuansheng Luo;Yong Gu;Jianjian Yan;Li Yao;Xinwu Li","doi":"10.1109/TETCI.2025.3641985","DOIUrl":"https://doi.org/10.1109/TETCI.2025.3641985","url":null,"abstract":"Converting face photos into sketch images has diverse applications in law enforcement and digital entertainment. However, this task remains challenging due to the significant modality differences between face photos and sketches. Existing methods often fail to simultaneously capture global and local features, and the synthesized face sketch images suffer from detail loss and style distortion. To address these limitations, a dual-branch network with Mamba-Transformer encoder and style-enhanced Wavelet convolution (SEWC) module is proposed for face sketch synthesis. Specifically, the Mamba-Transformer encoder based on a MambaVision block and a simplified variable-size window attention (SVSA) block is designed to efficiently capture both global and local features, enhancing texture preservation in the generated face sketches. Additionally, the SEWC module consisting of a style-enhanced convolution (SEConv) block and a Wavelet transform convolution (WTConv) block is built to refine style feature extraction and mitigate style distortion. Furthermore, within the SEWC module, an adaptive fusion module (AFM) is constructed to adaptively integrate shallow and deep style features through learnable weights.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"10 2","pages":"1454-1468"},"PeriodicalIF":6.0,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148509660","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Multi-Label Feature Selection by Maximum Affinity-Relevance, Maximum Affinity-Interaction and Minimum Affinity-Redundancy 基于最大亲和关联、最大亲和交互和最小亲和冗余的多标签特征选择

IF 6 3区计算机科学

IEEE Transactions on Emerging Topics in Computational Intelligence Pub Date : 2026-04-01 Epub Date: 2026-01-16 DOI: 10.1109/TETCI.2026.3650841

Jianhua Dai;Duo Xu

{"title":"Multi-Label Feature Selection by Maximum Affinity-Relevance, Maximum Affinity-Interaction and Minimum Affinity-Redundancy","authors":"Jianhua Dai;Duo Xu","doi":"10.1109/TETCI.2026.3650841","DOIUrl":"https://doi.org/10.1109/TETCI.2026.3650841","url":null,"abstract":"Multi-label feature selection solves the high-dimensional challenge problem in multi-label learning, and is widely used in pattern recognition, machine learning, and other related fields. However, most existing multi-label learning methods neglect the affinity between the features and the labels, which means the selected features may only provide categorization information for some of the labels rather than for all of them, resulting in one portion of labels obtaining sufficient categorization information, while another portion of labels obtains insufficient categorization information. In order to address this issue, we propose a new method for multi-label feature selection. Firstly, we propose the notion of affinity, which states that features are biased to provide categorical information for labels. Secondly, we propose three new feature evaluation metrics, affinity-relevance, affinity-redundancy, and affinity-interaction, which evaluate features in three dimensions from an affinity perspective. Thirdly, a new feature evaluation function called feature affinity significance, as well as a feature selection algorithm based on MARAImAR are proposed. Finally, experimental results on 15 multi-label benchmark datasets demonstrate that the proposed method outperforms the other nine representative multi-label feature selection methods.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"10 2","pages":"1805-1820"},"PeriodicalIF":6.0,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148510434","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Multi-Instance Multi-Label Learning via Label Specific Features and Label Enhancement with Collaborative Multi Objective Particle Swarm Optimization 基于标签特定特征的多实例多标签学习和基于协同多目标粒子群优化的标签增强

IF 6 3区计算机科学

IEEE Transactions on Emerging Topics in Computational Intelligence Pub Date : 2026-04-01 Epub Date: 2025-10-31 DOI: 10.1109/TETCI.2025.3623235

Xiang Bao;Fei Han;Qing-Hua Ling;Henry Han

{"title":"Multi-Instance Multi-Label Learning via Label Specific Features and Label Enhancement with Collaborative Multi Objective Particle Swarm Optimization","authors":"Xiang Bao;Fei Han;Qing-Hua Ling;Henry Han","doi":"10.1109/TETCI.2025.3623235","DOIUrl":"https://doi.org/10.1109/TETCI.2025.3623235","url":null,"abstract":"Multi-Instance Multi-Label (MIML) approaches predominantly rely on utilizing raw features to construct classifiers, which often struggle to effectively discern patterns and regularities within the data. Furthermore, these methods typically adopt a uniform label distribution assumption, overlooking the fact that importance of different labels often varies in real scenarios. To address these limitations, this paper proposes a MIML algorithm based on label-specific feature generation and label enhancement (MIML-LSFLE). Specifically, the proposed algorithm leverages a collaborative Multi-Objective Particle Swarm Optimization (MOPSO) to facilitate the selection and embedding of representative elements from the candidate prototype, thereby generates a tailored set of features for each pair of labels, classifiers are constructed based on bi-label specific features afterwards, and the collaborative MOPSO leverages a collaborative evolution strategy to maximize information exchange and sharing between different populations during the selection of personal best solution (Pbest). Subsequently, the algorithm exploits the topological information within the feature space and the inter-label correlations during the process of label enhancement (LE) where a comprehensive iterative approach is employed. Ultimately, the recovered label distribution is integrated into the label prediction phase through experience-weighted voting. The experimental results demonstrate that the proposed method achieves competitive performance compared with the classical and state-of-the-art MIML models from the perspective of several classification indicators.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"10 2","pages":"1779-1793"},"PeriodicalIF":6.0,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148511350","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

An Enhanced Policy-Based Reinforcement Learning With Hierarchical Reward for Dynamic No-Wait Flowshop Scheduling Problem 动态无等待流水车间调度问题的一种改进的分层奖励策略强化学习

IF 6 3区计算机科学

IEEE Transactions on Emerging Topics in Computational Intelligence Pub Date : 2026-04-01 Epub Date: 2025-11-12 DOI: 10.1109/TETCI.2025.3628758

Junyi Du;Ling Wang;Rui Li

{"title":"An Enhanced Policy-Based Reinforcement Learning With Hierarchical Reward for Dynamic No-Wait Flowshop Scheduling Problem","authors":"Junyi Du;Ling Wang;Rui Li","doi":"10.1109/TETCI.2025.3628758","DOIUrl":"https://doi.org/10.1109/TETCI.2025.3628758","url":null,"abstract":"Machine learning has demonstrated remarkable effectiveness in solving scheduling problems through end-to-end optimization. However, dynamic events introduce uncertainty and pose significant challenges to scheduling performance. In this paper, we address the No-Wait Flowshop Scheduling Problem with dynamic job arrivals by proposing a novel generative scheduling algorithm. The problem is formulated as a Markov Decision Process and decomposed into single-step scheduling decisions, significantly reducing the complexity of global optimization and computational overhead. Our algorithm leverages an attention-based neural network to efficiently capture both individual task features and global task dependencies, enabling adaptive and effective generative scheduling in dynamic environments. A key innovation of our approach lies in the attention-based encoder-decoder architecture, where the encoder precisely extracts task-specific features and global interrelationships, while the decoder dynamically generates scheduling decisions through an adaptive query-driven mechanism, ensuring flexibility and real-time responsiveness. In addition, we propose a two-stage knowledge-guided reward mechanism to enhance training efficiency. In the initial stage, the reward system focuses on minimizing local idle times, while in the subsequent stage it shifts focus to optimizing the overall makespan objective. Extensive experimental evaluations demonstrate that the proposed algorithm consistently outperforms traditional scheduling rules and reinforcement learning-based methods, achieving superior performance in minimizing makespan across various problem scales.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"10 2","pages":"1271-1283"},"PeriodicalIF":6.0,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148512143","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0