Journal of Computational Science最新文献

Pattern embedding driven lightweight quadrilateral mesh generation for engineering applications 面向工程应用的模式嵌入驱动轻量级四边形网格生成

IF 3.7 3区计算机科学

Journal of Computational Science Pub Date : 2026-04-01 Epub Date: 2026-01-27 DOI: 10.1016/j.jocs.2026.102790

Junxian Liu , Minglin Li , Wei Ye , Jinxu Liu , Lianfeng Lai

{"title":"Pattern embedding driven lightweight quadrilateral mesh generation for engineering applications","authors":"Junxian Liu , Minglin Li , Wei Ye , Jinxu Liu , Lianfeng Lai","doi":"10.1016/j.jocs.2026.102790","DOIUrl":"10.1016/j.jocs.2026.102790","url":null,"abstract":"<div><div>Current mainstream engineering simulation software struggles to generate high-quality quadrilateral meshes during the preprocessing stage, and achieving such meshes typically relies on numerical solvers. To enable efficient and high-quality quadrilateral mesh generation for complex geometries in open-source environments, this paper proposes a lightweight quadrilateral mesh reconstruction method. The method first takes a triangular mesh model as input and parameterizes it onto a 2D plane and performs polygonal layout decomposition using directional constraints. A greedy propagation strategy is then employed to assign uniform subdivision parameters to layout boundaries, replacing global integer programming. Finally, quadrilateral mesh templates are embedded to complete the mesh generation. The entire process avoids reliance on numerical solvers. Experiments on publicly available benchmark models demonstrate that, even without using solvers, the generated meshes achieve quality comparable to or better than mainstream approaches, providing a low-cost, high-accuracy preprocessing solution for engineering simulations.</div></div>","PeriodicalId":48907,"journal":{"name":"Journal of Computational Science","volume":"95 ","pages":"Article 102790"},"PeriodicalIF":3.7,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146057291","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

Adaptive discount factor for accelerating policy learning considering long-term returns in reinforcement learning with non-stationary environments 非平稳环境下考虑长期收益的强化学习加速策略学习的自适应折现因子

IF 3.7 3区计算机科学

Journal of Computational Science Pub Date : 2026-04-01 Epub Date: 2026-02-02 DOI: 10.1016/j.jocs.2026.102787

Kazuki Ogawa , Takeru Goto , Tatsuhito Aihara , Gaku Minorikawa

{"title":"Adaptive discount factor for accelerating policy learning considering long-term returns in reinforcement learning with non-stationary environments","authors":"Kazuki Ogawa , Takeru Goto , Tatsuhito Aihara , Gaku Minorikawa","doi":"10.1016/j.jocs.2026.102787","DOIUrl":"10.1016/j.jocs.2026.102787","url":null,"abstract":"<div><div>This paper proposes a method for adaptively controlling the discount factor based on the loss function value, aiming to improve learning speed and stability while accounting for long-term returns in both stationary and non-stationary environments. The discount factor determines the weighting of future rewards, presenting a trade-off: higher values prioritize long-term returns, though they are more likely to result in instability and slower convergence. Lower values facilitate faster learning at the cost of smaller cumulative rewards, even when learning has converged. To address this dilemma, dynamically adjusting the discount factor based on the progress of learning allows the agent to prioritize short-term rewards in the early stages while shifting toward long-term rewards as learning progresses. Also, the environment is not always stationary; real-world applications, such as machine control, often involve non-stationary environments where conditions change over time due to factors like aging. In the proposed method, the loss function value serves as an indicator of learning progress and is treated as the control variable for PID-based adjustment of the discount factor. The key feature of this method is the simplification of controller parameter design using a model matching approach to determine the PID parameters and simplifies parameter tuning. When the environment changes further, lowering the discount factor allows for quicker adaptation to the changes in the environment. Through simulation experiments, we demonstrate that the proposed method outperforms conventional approaches in terms of both learning speed and stability.</div></div>","PeriodicalId":48907,"journal":{"name":"Journal of Computational Science","volume":"95 ","pages":"Article 102787"},"PeriodicalIF":3.7,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146191177","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

Classification of piano performers with deep learning models 用深度学习模型对钢琴演奏者进行分类

IF 3.7 3区计算机科学

Journal of Computational Science Pub Date : 2026-04-01 Epub Date: 2026-02-02 DOI: 10.1016/j.jocs.2026.102804

Jan Mycka , Jacek Mańdziuk

{"title":"Classification of piano performers with deep learning models","authors":"Jan Mycka , Jacek Mańdziuk","doi":"10.1016/j.jocs.2026.102804","DOIUrl":"10.1016/j.jocs.2026.102804","url":null,"abstract":"<div><div>Artists’ classification based on their unique performing style remains a rarely explored problem in music information retrieval. Nevertheless, it is an extremely intriguing subject that challenges even experienced experts. In this study, deep learning approaches, namely Convolutional Recurrent Neural Network (CRNN) and Transformer-based model, are employed to classify piano performers based on audio recordings. CRNN combines convolutional layers for the extraction of spectral characteristics with recurrent layers for temporal sequence modeling whereas Transformer-based models are widely used for sequence analysis, which makes both architectures suitable for music processing. To evaluate the ability of the model to recognize performers, multiple tests are conducted, with a varying number of both performers and composers of performed music, assessing how these variations impact accuracy. Depending on these aspects, the models’ accuracy ranges from <span><math><mrow><mo>∼</mo><mn>70</mn><mtext>%</mtext></mrow></math></span> to <span><math><mrow><mo>∼</mo><mn>98</mn><mtext>%</mtext></mrow></math></span>. The findings show that pianist classification using deep learning is feasible, but could still be improved with further refinements. The potential applications include digital music archiving, historical performance analysis, or AI-assisted stylistic studies.</div></div>","PeriodicalId":48907,"journal":{"name":"Journal of Computational Science","volume":"95 ","pages":"Article 102804"},"PeriodicalIF":3.7,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146191333","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

Discovery of Bus Loop Scheduling strategies with reinforcement learning to minimize commuters’ waiting and travel times 基于强化学习的公交环路调度策略的发现，以最大限度地减少通勤者的等待和旅行时间

IF 3.7 3区计算机科学

Journal of Computational Science Pub Date : 2026-04-01 Epub Date: 2026-02-11 DOI: 10.1016/j.jocs.2026.102807

Andri Pradana, Lock Yue Chew

{"title":"Discovery of Bus Loop Scheduling strategies with reinforcement learning to minimize commuters’ waiting and travel times","authors":"Andri Pradana, Lock Yue Chew","doi":"10.1016/j.jocs.2026.102807","DOIUrl":"10.1016/j.jocs.2026.102807","url":null,"abstract":"<div><div>In this paper, we investigate the application of two reinforcement learning methods, known as the Dueling Double Deep Q-Network and Soft Actor–Critic to discover bus scheduling strategies and compare them against conventional approaches. In particular, we look into real-time control strategies where buses may choose to stay or leave at bus stops. We explore both waiting time and travel time as the optimization objectives. The results for uniform bus frequency show that average waiting time can be reduced by allowing buses to stay longer at stops with higher passengers’ arrival rate but at the cost of increased average travel time. This is also supported by our analytical calculation on a theoretical bus loop model. We then apply our method to a model based on a real world bus loop in Nanyang Technological University. The results highlight the potential benefit of reinforcement learning methods to find novel strategies that can be better than conventional approaches. The similar performance of the two distinct reinforcement learning methods also serves as independent verification of the validity of the strategies obtained. This is an extended version of our ICCS 2025 conference paper “Bus Loop Scheduling with Dueling Double Deep Q Network” Pradana and Chew (2025) <span><span>[1]</span></span>, with the main addition of the application of the Soft Actor–Critic method which has to be modified to handle the optimization problem described in this paper.</div></div>","PeriodicalId":48907,"journal":{"name":"Journal of Computational Science","volume":"95 ","pages":"Article 102807"},"PeriodicalIF":3.7,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146191175","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

Machine learning in proton exchange membrane fuel cell studies: Nafion 112 membrane 质子交换膜燃料电池研究中的机器学习：Nafion 112膜

IF 3.7 3区计算机科学

Journal of Computational Science Pub Date : 2026-04-01 Epub Date: 2026-02-07 DOI: 10.1016/j.jocs.2026.102802

A.L. De Bortoli

引用次数: 0

A novel edge reconstruction strategy for command and control networks: Balancing shortest path length and node importance 一种新的指挥与控制网络边缘重建策略：平衡最短路径长度和节点重要性

IF 3.7 3区计算机科学

Journal of Computational Science Pub Date : 2026-04-01 Epub Date: 2026-01-28 DOI: 10.1016/j.jocs.2026.102801

Xiue Gao , Lingdong Sun , Yufeng Chen , Guimei Pang , Bo Chen , Zhengtao Xiang

{"title":"A novel edge reconstruction strategy for command and control networks: Balancing shortest path length and node importance","authors":"Xiue Gao , Lingdong Sun , Yufeng Chen , Guimei Pang , Bo Chen , Zhengtao Xiang","doi":"10.1016/j.jocs.2026.102801","DOIUrl":"10.1016/j.jocs.2026.102801","url":null,"abstract":"<div><div>Addressing the deficiencies prevalent in current edge reconstruction methodologies within command and control networks, characterized by overly uniform degree distributions and inadequate treatment of isolated nodes, this paper introduces a novel approach that integrates considerations of both shortest path length and node importance. Firstly, we delineate the initial load and capacity of edges through the incorporation of edge importance and edge hierarchy. Subsequently, node importance is defined by leveraging node betweenness centrality and node degree. Through the amalgamation of shortest path length and node importance, we devise a methodology for computing the edge reconstruction index. Following this, we prioritize the establishment of new edges based on the maximum edge reconstruction index and devise a neighbor load allocation strategy grounded on remaining capacity. Finally, simulation experiments are conducted to compare the global efficiency, reconstruction efficiency, reconstruction cost, and network connectivity coefficient of various reconstruction strategies. The results showcase a substantial reduction in reconstruction cost and a notable enhancement in reconstruction efficiency with the proposed methodology.</div></div>","PeriodicalId":48907,"journal":{"name":"Journal of Computational Science","volume":"95 ","pages":"Article 102801"},"PeriodicalIF":3.7,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146191332","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

dpGMM: A new R package for efficient and robust Gaussian mixture modeling of 1D and 2D data dpGMM：一个新的R包，用于高效和鲁棒的一维和二维数据高斯混合建模

IF 3.7 3区计算机科学

Journal of Computational Science Pub Date : 2026-04-01 Epub Date: 2026-02-11 DOI: 10.1016/j.jocs.2026.102811

Joanna Zyla , Kamila Szumala , Andrzej Polanski , Joanna Polanska , Michal Marczyk

{"title":"dpGMM: A new R package for efficient and robust Gaussian mixture modeling of 1D and 2D data","authors":"Joanna Zyla , Kamila Szumala , Andrzej Polanski , Joanna Polanska , Michal Marczyk","doi":"10.1016/j.jocs.2026.102811","DOIUrl":"10.1016/j.jocs.2026.102811","url":null,"abstract":"<div><div>Gaussian Mixture Modeling (GMM) is a powerful clustering and density estimation method with various applications in data analysis. We introduce an R package, <em>dpGMM</em>, a complete set of tools/procedures to analyze 1D or 2D data (binned or continuous), including the most efficient existing solutions to problems of fitting GMM to data by the recursive expectation-maximization (EM) algorithm. The effectiveness of the <em>dpGMM</em> package comes from leveraging the power of EM recursions by: (i) precise choice of the initial mixture parameters obtained with the use of the dynamic programming-based partition of data, and (ii) augmenting each M step with additional conditions aimed at preventing instability/divergence and accelerating the rate of convergence of iterations. <em>dpGMM</em> is implemented as a wrapper that allows for searching the best decomposition in the scenario with an unknown number of Gaussian mixture components, by using various information criteria, as well as with a fixed number of components. We compared <em>dpGMM</em> with three other R packages using synthetic and real biological datasets, performing large-scale computations to assess the performance of these GMM implementations across various scenarios.</div></div>","PeriodicalId":48907,"journal":{"name":"Journal of Computational Science","volume":"95 ","pages":"Article 102811"},"PeriodicalIF":3.7,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146191176","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

PPCN: Identifying influential nodes based on propagation probability model considering neighbors in complex networks PPCN：复杂网络中基于考虑邻居的传播概率模型的影响节点识别

IF 3.7 3区计算机科学

Journal of Computational Science Pub Date : 2026-04-01 Epub Date: 2026-02-05 DOI: 10.1016/j.jocs.2026.102792

Bo Liu , Juncan Wei , Kaile Chen , Jinfeng Gao , Mengqi Xue , Ping Wu

{"title":"PPCN: Identifying influential nodes based on propagation probability model considering neighbors in complex networks","authors":"Bo Liu , Juncan Wei , Kaile Chen , Jinfeng Gao , Mengqi Xue , Ping Wu","doi":"10.1016/j.jocs.2026.102792","DOIUrl":"10.1016/j.jocs.2026.102792","url":null,"abstract":"<div><div>In network science, where identifying influential nodes is crucial for understanding network structures and functions, propagation probability model often is an effective approach. However, most of existing propagation-based methods often overlook propagation via non-shortest paths; moreover, they ignore that nodes with equal propagation probability can exert different influence due to inherent differences in node quality. In this paper, we propose the Propagation Probability Model Considering Neighbors (PPCN) for identifying influential nodes. The method considers these two aspects. It calculates node importance through the inherent influence of the node and the influence transmitted to it by other nodes. Specifically, we compute a more effective propagation probability between node pairs by refining propagation paths, integrating it with node’s inherent influence to represent its transmitted influence. The node’s inherent influence is determined by the sum of its degree value and neighbor diversity. To evaluate the performance of the PPCN method, we compared it with ten common algorithms on nine real-world networks, assessing four aspects via the susceptible-infected-recovered (SIR) model, and the experimental results demonstrate that the proposed algorithm exhibits good performance in most cases.</div></div>","PeriodicalId":48907,"journal":{"name":"Journal of Computational Science","volume":"95 ","pages":"Article 102792"},"PeriodicalIF":3.7,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146191173","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

GRASFormer: A computational framework for gradient-regularized and entropy-stabilized multi-task transformer optimization GRASFormer：一个梯度正则化和熵稳定的多任务变压器优化计算框架

IF 3.7 3区计算机科学

Journal of Computational Science Pub Date : 2026-04-01 Epub Date: 2026-02-03 DOI: 10.1016/j.jocs.2026.102805

Pulkit Dwivedi , Benazir Islam

{"title":"GRASFormer: A computational framework for gradient-regularized and entropy-stabilized multi-task transformer optimization","authors":"Pulkit Dwivedi , Benazir Islam","doi":"10.1016/j.jocs.2026.102805","DOIUrl":"10.1016/j.jocs.2026.102805","url":null,"abstract":"<div><div>Multi-task learning (MTL) with transformer architectures represents a computationally powerful paradigm for joint optimization across related objectives. However, the underlying optimization dynamics remain difficult to stabilize due to inter-task gradient conflicts and excessively sharp attention landscapes, which lead to high-curvature loss surfaces and degraded generalization. This paper introduces <em>GRASFormer</em>, a Gradient Regularized Attention Sharpness Transformer, as a unified computational framework that systematically addresses these challenges through dual regularization. GRASFormer models the interdependence among task gradients via pairwise cosine similarity and dynamically reprojects conflicting directions to achieve smoother multi-objective descent. Concurrently, an entropy-based attention regularizer is formulated to control attention sharpness, thereby ensuring numerically stable and broadly distributed focus across transformer heads. The resulting framework constitutes a generalizable computational model for stable and efficient optimization in multi-task deep learning. Comprehensive experiments on synthetic optimization landscapes and benchmark datasets, including MultiMNIST, FashionMNIST-Edge, CelebA, and SVHN-Length, demonstrate that GRASFormer achieves faster convergence, improved accuracy, and enhanced robustness compared to existing state-of-the-art MTL optimizers. Statistical analyses and ablation diagnostics further confirm the theoretical consistency and computational efficiency of the proposed approach.</div></div>","PeriodicalId":48907,"journal":{"name":"Journal of Computational Science","volume":"95 ","pages":"Article 102805"},"PeriodicalIF":3.7,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146191334","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

Built-in look-ahead to counter reaction delay in realistic traffic flow simulation 内置前瞻性，以应对现实交通流模拟的反应延迟

IF 3.7 3区计算机科学

Journal of Computational Science Pub Date : 2026-04-01 Epub Date: 2026-02-02 DOI: 10.1016/j.jocs.2026.102803

Daiheng NI

引用次数: 0