Pub Date : 2025-12-25DOI: 10.1016/j.ins.2025.123033
Weronika Smolak-Dyżewska , Dawid Malarz , Kornel Howil , Jan Kaczmarczyk , Marcin Mazur , Przemysław Spurek
Implicit Neural Representations (INRs) employ neural networks to approximate discrete data as continuous functions. In the context of video data, such models can be utilized to transform the coordinates of pixel locations along with frame occurrence times (or indices) into RGB color values. Although INRs facilitate effective compression, they are unsuitable for editing purposes. One potential solution is to use a 3D Gaussian Splatting (3DGS) based model, such as the Video Gaussian Representation (VGR), which is capable of encoding video as a multitude of 3D Gaussians and is applicable for numerous video processing operations, including editing. Nevertheless, in this case, the capacity for modification is constrained to a limited set of basic transformations. To address this issue, we introduce the Video Gaussian Splatting (VeGaS) model, which enables complex modifications of video data. To construct VeGaS, we propose a novel family of Folded-Gaussian distributions designed to capture nonlinear dynamics in a video stream and model consecutive frames by 2D Gaussians obtained as respective conditional distributions. Our experiments demonstrate that VeGaS outperforms state-of-the-art solutions in frame reconstruction tasks and allows expressive modifications of video data. The code is available at: https://github.com/gmum/VeGaS.
{"title":"VeGaS: Video Gaussian Splatting","authors":"Weronika Smolak-Dyżewska , Dawid Malarz , Kornel Howil , Jan Kaczmarczyk , Marcin Mazur , Przemysław Spurek","doi":"10.1016/j.ins.2025.123033","DOIUrl":"10.1016/j.ins.2025.123033","url":null,"abstract":"<div><div>Implicit Neural Representations (INRs) employ neural networks to approximate discrete data as continuous functions. In the context of video data, such models can be utilized to transform the coordinates of pixel locations along with frame occurrence times (or indices) into RGB color values. Although INRs facilitate effective compression, they are unsuitable for editing purposes. One potential solution is to use a 3D Gaussian Splatting (3DGS) based model, such as the Video Gaussian Representation (VGR), which is capable of encoding video as a multitude of 3D Gaussians and is applicable for numerous video processing operations, including editing. Nevertheless, in this case, the capacity for modification is constrained to a limited set of basic transformations. To address this issue, we introduce the Video Gaussian Splatting (VeGaS) model, which enables complex modifications of video data. To construct VeGaS, we propose a novel family of Folded-Gaussian distributions designed to capture nonlinear dynamics in a video stream and model consecutive frames by 2D Gaussians obtained as respective conditional distributions. Our experiments demonstrate that VeGaS outperforms state-of-the-art solutions in frame reconstruction tasks and allows expressive modifications of video data. The code is available at: <span><span>https://github.com/gmum/VeGaS</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"735 ","pages":"Article 123033"},"PeriodicalIF":6.8,"publicationDate":"2025-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145886517","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-25DOI: 10.1016/j.ins.2025.123035
Genqiang Wu , Xianyao Xia , Yeping He
Current differential privacy frameworks face significant challenges: vulnerability to correlated data attacks and suboptimal utility-privacy tradeoffs. To address these limitations, we establish a novel information-theoretic foundation for Dalenius’ privacy vision using Shannon’s perfect secrecy framework. By leveraging the fundamental distinction between cryptographic systems (small secret keys) and privacy mechanisms (massive datasets), we replace differential privacy’s restrictive independence assumption with practical partial knowledge constraints ().
We propose an information privacy framework achieving Dalenius security with quantifiable utility-privacy tradeoffs. Crucially, we prove that foundational mechanisms—random response, exponential, and Gaussian channels—satisfy Dalenius’ requirements while preserving group privacy and composition properties. Our channel capacity analysis reduces infinite-dimensional evaluations to finite convex optimizations, enabling direct application of information-theoretic tools.
Empirical evaluation demonstrates that individual channel capacity (maximal information leakage of each individual) decreases with increasing entropy constraint , and our framework achieves superior utility-privacy tradeoffs compared to classical differential privacy mechanisms under equivalent privacy guarantees. The framework is extended to computationally bounded adversaries via Yao’s theory, unifying cryptographic and statistical privacy paradigms. Collectively, these contributions provide a theoretically grounded path toward practical, composable privacy—subject to future resolution of the tradeoff characterization—with enhanced resilience to correlation attacks.
{"title":"Achieving Dalenius’ goal of data privacy with practical assumptions","authors":"Genqiang Wu , Xianyao Xia , Yeping He","doi":"10.1016/j.ins.2025.123035","DOIUrl":"10.1016/j.ins.2025.123035","url":null,"abstract":"<div><div>Current differential privacy frameworks face significant challenges: vulnerability to correlated data attacks and suboptimal utility-privacy tradeoffs. To address these limitations, we establish a novel information-theoretic foundation for Dalenius’ privacy vision using Shannon’s perfect secrecy framework. By leveraging the fundamental distinction between cryptographic systems (small secret keys) and privacy mechanisms (massive datasets), we replace differential privacy’s restrictive independence assumption with practical partial knowledge constraints (<span><math><mi>H</mi><mo>(</mo><mi>X</mi><mo>)</mo><mo>≥</mo><mi>b</mi></math></span>).</div><div>We propose an information privacy framework achieving Dalenius security with quantifiable utility-privacy tradeoffs. Crucially, we prove that foundational mechanisms—random response, exponential, and Gaussian channels—satisfy Dalenius’ requirements while preserving group privacy and composition properties. Our channel capacity analysis reduces infinite-dimensional evaluations to finite convex optimizations, enabling direct application of information-theoretic tools.</div><div>Empirical evaluation demonstrates that individual channel capacity (maximal information leakage of each individual) decreases with increasing entropy constraint <span><math><mi>b</mi></math></span>, and our framework achieves superior utility-privacy tradeoffs compared to classical differential privacy mechanisms under equivalent privacy guarantees. The framework is extended to computationally bounded adversaries via Yao’s theory, unifying cryptographic and statistical privacy paradigms. Collectively, these contributions provide a theoretically grounded path toward practical, composable privacy—subject to future resolution of the tradeoff characterization—with enhanced resilience to correlation attacks.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"735 ","pages":"Article 123035"},"PeriodicalIF":6.8,"publicationDate":"2025-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145852496","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-24DOI: 10.1016/j.ins.2025.123024
Zhan ao Huang , Canghong Shi , Jia He , Xiaojie Li , Xi Wu
Oversampling is a classical method used by neural networks to address the learning bias caused by imbalanced data learning. However, most oversampling approaches are limited to the empirical distribution of known data, disregarding its insufficient representation problem. Consequently, the learning bias cannot be fully alleviated. In this paper, a novel sampling paradigm outside the empirical distribution is developed to address imbalanced data classification with an insufficient empirical distribution. Specifically, we sample absent minority samples that have low majority probability attributes outside the empirical distribution by using normalizing flow technology and surrogate complement set sampling. A sampling space of absent minority samples is constructed by combining different stages of the generation direction of the normalizing flow model. In addition, to preserve the details of adjacent areas between classes, we transform the sampling constraint from global probability to local cluster distance. Alleviating the insufficient empirical distribution by incorporating the absent minority samples in neural network optimization. We validated the proposed method on KEEL imbalanced datasets and application tests. The proposed method shows obvious advantages over the state-of-the-art absent minority oversampling technologies.
{"title":"AMOS: Absent minority oversampling neural network for imbalanced data classification","authors":"Zhan ao Huang , Canghong Shi , Jia He , Xiaojie Li , Xi Wu","doi":"10.1016/j.ins.2025.123024","DOIUrl":"10.1016/j.ins.2025.123024","url":null,"abstract":"<div><div>Oversampling is a classical method used by neural networks to address the learning bias caused by imbalanced data learning. However, most oversampling approaches are limited to the empirical distribution of known data, disregarding its insufficient representation problem. Consequently, the learning bias cannot be fully alleviated. In this paper, a novel sampling paradigm outside the empirical distribution is developed to address imbalanced data classification with an insufficient empirical distribution. Specifically, we sample absent minority samples that have low majority probability attributes outside the empirical distribution by using normalizing flow technology and surrogate complement set sampling. A sampling space of absent minority samples is constructed by combining different stages of the generation direction of the normalizing flow model. In addition, to preserve the details of adjacent areas between classes, we transform the sampling constraint from global probability to local cluster distance. Alleviating the insufficient empirical distribution by incorporating the absent minority samples in neural network optimization. We validated the proposed method on KEEL imbalanced datasets and application tests. The proposed method shows obvious advantages over the state-of-the-art absent minority oversampling technologies.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"735 ","pages":"Article 123024"},"PeriodicalIF":6.8,"publicationDate":"2025-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145852494","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-24DOI: 10.1016/j.ins.2025.123023
Lirong Zhang , Yi Lin , Suwan Yin , Wu Deng , Hongyu Yang , Jianwei Zhang
The airport field resource scheduling problem (AFRS) plays a crucial role in airport surface operations but is typically formulated as separate optimization tasks, leading to inefficient airport operations. To bridge this gap, a new problem model is proposed to implement intra-resource collaboration based on the attributes of airport operation to enhance operational flexibility, namely the joint scheduling model of runway and taxiway considering uncertain taxiing time (RTU). To reduce flight delays caused by deviations from the estimated schedule time, a two-stage queue completion mechanism is proposed to adjust runway sequencing by considering both the separation constraints and their estimated scheduled times in the flight plan. In addition, to alleviate the high costs associated with frequent taxiing conflicts, a time-shifting mechanism is proposed to transfer time costs based on flight priority, thereby enhancing taxiing efficiency and reducing conflict propagation. In this work, an improved ant colony optimization (ACO) algorithm, namely an ACO algorithm based on Metropolis- and screening- strategies (MS-ACO), is proposed to solve the RTU model. To address the local optimum problem, a Metropolis strategy is proposed to guide the optimization direction. To cope with the route-point skipping, path circuitousness, and routing errors caused by the randomness of the ACO algorithm, a screening strategy is designed to recalibrate taxiing paths based on the physical constraints of the route network. A real-world dataset is applied to validate the proposed model and solution method. The experimental results indicate that the proposed RTU model outperforms the separate optimization model, with delay time decreasing by approximately 8.4 , thereby identifying optimal sequencing and taxiing plans that significantly reduce conflicts and delays. Most importantly, the generalizability of the RTU model is confirmed through comparisons with other competitive methods, and the efficiency and effectiveness of the MS-ACO algorithm are validated against state-of-the-art algorithms. The proposed multi-resource optimization framework can be applied to enable integrated analysis of runway and taxiway operations, thereby supporting airport operations.
机场现场资源调度问题（AFRS）在机场地面运行中起着至关重要的作用，但通常被制定为单独的优化任务，导致机场运行效率低下。为了弥补这一缺陷，提出了一种基于机场运行属性的资源内协同问题模型，即考虑不确定滑行时间（RTU）的跑道和滑道联合调度模型。为了减少因偏离预计计划时间而造成的航班延误，提出了一种两阶段排队完成机制，同时考虑飞行计划中的分离约束和预计计划时间，对跑道排序进行调整。此外，针对滑行冲突频繁带来的高成本问题，提出了一种基于飞行优先级的时间转移机制来转移时间成本，从而提高滑行效率，减少冲突传播。本文提出了一种改进的蚁群优化算法，即基于Metropolis- and - screening- strategies的蚁群优化算法（MS-ACO）来求解RTU模型。为了解决局部最优问题，提出了Metropolis策略来指导优化方向。针对蚁群算法的随机性导致的路径点跳变、路径迂回和路由错误等问题，设计了一种基于路由网络物理约束的滑行路径重新标定策略。应用一个实际数据集验证了所提出的模型和求解方法。实验结果表明，所提RTU模型优于单独优化模型，延迟时间减少了约8.4%，从而确定了最优的排序和滑行计划，显著减少了冲突和延迟。最重要的是，通过与其他竞争方法的比较，验证了RTU模型的可泛化性，并与最先进的算法对比验证了MS-ACO算法的效率和有效性。建议的多资源优化框架可应用于跑道和滑行道运营的综合分析，从而支持机场运营。
{"title":"Joint scheduling of runway and taxiway considering uncertain taxiing time based on an improved ACO algorithm","authors":"Lirong Zhang , Yi Lin , Suwan Yin , Wu Deng , Hongyu Yang , Jianwei Zhang","doi":"10.1016/j.ins.2025.123023","DOIUrl":"10.1016/j.ins.2025.123023","url":null,"abstract":"<div><div>The airport field resource scheduling problem (AFRS) plays a crucial role in airport surface operations but is typically formulated as separate optimization tasks, leading to inefficient airport operations. To bridge this gap, a new problem model is proposed to implement intra-resource collaboration based on the attributes of airport operation to enhance operational flexibility, namely the joint scheduling model of <u>r</u>unway and <u>t</u>axiway considering <u>u</u>ncertain taxiing time (RTU). To reduce flight delays caused by deviations from the estimated schedule time, a two-stage queue completion mechanism is proposed to adjust runway sequencing by considering both the separation constraints and their estimated scheduled times in the flight plan. In addition, to alleviate the high costs associated with frequent taxiing conflicts, a time-shifting mechanism is proposed to transfer time costs based on flight priority, thereby enhancing taxiing efficiency and reducing conflict propagation. In this work, an improved ant colony optimization (ACO) algorithm, namely an ACO algorithm based on <u>M</u>etropolis- and <u>s</u>creening- strategies (MS-ACO), is proposed to solve the RTU model. To address the local optimum problem, a Metropolis strategy is proposed to guide the optimization direction. To cope with the route-point skipping, path circuitousness, and routing errors caused by the randomness of the ACO algorithm, a screening strategy is designed to recalibrate taxiing paths based on the physical constraints of the route network. A real-world dataset is applied to validate the proposed model and solution method. The experimental results indicate that the proposed RTU model outperforms the separate optimization model, with delay time decreasing by approximately 8.4 <span><math><mrow><mi>%</mi></mrow></math></span>, thereby identifying optimal sequencing and taxiing plans that significantly reduce conflicts and delays. Most importantly, the generalizability of the RTU model is confirmed through comparisons with other competitive methods, and the efficiency and effectiveness of the MS-ACO algorithm are validated against state-of-the-art algorithms. The proposed multi-resource optimization framework can be applied to enable integrated analysis of runway and taxiway operations, thereby supporting airport operations.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"735 ","pages":"Article 123023"},"PeriodicalIF":6.8,"publicationDate":"2025-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145886516","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-24DOI: 10.1016/j.ins.2025.123031
Pengsong Zhang , Mi Wen , Zhou Zhu , Dongyang Li
Real-world data often exhibit severe class imbalance with long-tail distribution, presenting significant challenges for classification tasks, especially in federated learning (FL). Existing research predominantly addresses the global distribution problem of long-tail data, aiming to improve the efficiency of the global model. However, these studies often overlook the potential presence of long-tail distributions at the local level and neglect performance optimization at this level, leading to degraded personalized performance and biased aggregation. This study introduces FedDream, a federated learning framework that consistently improves the performance of both local and global models. Specifically, a shared backbone network is employed to capture global trends, upon which a Dynamic Adaptive Equiangular Tight Frame (DA-ETF), inspired by neural collapse, is constructed to guide the backbone in dynamically learning balanced feature representations. Subsequently, we treat these backbone networks as expert models and train the personalized models through the Dynamic Class-Conditional Mixture of Experts (DCC-MoE). Finally, a global model robust to malicious attacks is obtained by hierarchically aggregating the personalized models. Comprehensive experimental results across CIFAR-10-LT, CIFAR-100-LT, and ImageNet-LT datasets demonstrate that FedDream significantly improves both accuracy and security across multiple benchmarks, offering a novel perspective for FL under long-tailed data distributions.
{"title":"Taming the long tail in federated learning: A unified global and personalized model framework","authors":"Pengsong Zhang , Mi Wen , Zhou Zhu , Dongyang Li","doi":"10.1016/j.ins.2025.123031","DOIUrl":"10.1016/j.ins.2025.123031","url":null,"abstract":"<div><div>Real-world data often exhibit severe class imbalance with long-tail distribution, presenting significant challenges for classification tasks, especially in federated learning (FL). Existing research predominantly addresses the global distribution problem of long-tail data, aiming to improve the efficiency of the global model. However, these studies often overlook the potential presence of long-tail distributions at the local level and neglect performance optimization at this level, leading to degraded personalized performance and biased aggregation. This study introduces FedDream, a federated learning framework that consistently improves the performance of both local and global models. Specifically, a shared backbone network is employed to capture global trends, upon which a Dynamic Adaptive Equiangular Tight Frame (DA-ETF), inspired by neural collapse, is constructed to guide the backbone in dynamically learning balanced feature representations. Subsequently, we treat these backbone networks as expert models and train the personalized models through the Dynamic Class-Conditional Mixture of Experts (DCC-MoE). Finally, a global model robust to malicious attacks is obtained by hierarchically aggregating the personalized models. Comprehensive experimental results across CIFAR-10-LT, CIFAR-100-LT, and ImageNet-LT datasets demonstrate that FedDream significantly improves both accuracy and security across multiple benchmarks, offering a novel perspective for FL under long-tailed data distributions.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"735 ","pages":"Article 123031"},"PeriodicalIF":6.8,"publicationDate":"2025-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145940326","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-23DOI: 10.1016/j.ins.2025.123015
N.R. Rejin Paul , V. Nallarasan , Nallam Krishnaiah , L. Guganathan
Problem statement
Cloud environments, while managing a huge volume of sensitive data, are increasingly susceptible to frequent and sophisticated cyber-attacks from interconnected IoT devices. The limitations of traditional intrusion detection due to class imbalance, high false alarm rates, and poor scalability reduce the reliability of intrusion detection in dynamic cloud infrastructures.
Proposed model
To address these, the paper proposes a Blockchain-Integrated IDS that contains an optimized Cosine Convolutional Neural Network enhanced with the Energy Valley Optimization Algorithm, CCNN-EVOA. The framework is designed for accurate detection with scalability and privacy-preserving data handling in real-time cloud environments.
Key methods
It integrates several components to achieve better performance of the IDS. A JDAGEM is used to carry out advanced pre-processing and feature extraction. Sensitive information is kept secure by PEDS. The Adaptive Blockchain Sharding Protocol allows for scalable, distributed, and tamper-resistant processing. The CCNN parameters are tuned using the EVOA algorithm for enhanced generalization and reduced false alarms.
Results
The framework is validated on real-world IoT-based attack scenarios using the BoT-IoT and UNSW-NB15 datasets. The proposed system has reached a high accuracy of 99.7% with notable improvements compared to state-of-the-art IDS techniques.
Implications
These findings affirm that Blockchain-Integrated IDS based on CCNN-EVOA provides a reliable, scalable, and privacy-preserving solution for contemporary IoT-enabled cloud infrastructures.
{"title":"Blockchain-integrated intrusion detection system with optimized cosine CNN for enhanced privacy and security in cloud computing","authors":"N.R. Rejin Paul , V. Nallarasan , Nallam Krishnaiah , L. Guganathan","doi":"10.1016/j.ins.2025.123015","DOIUrl":"10.1016/j.ins.2025.123015","url":null,"abstract":"<div><h3>Problem statement</h3><div>Cloud environments, while managing a huge volume of sensitive data, are increasingly susceptible to frequent and sophisticated cyber-attacks from interconnected IoT devices. The limitations of traditional intrusion detection due to class imbalance, high false alarm rates, and poor scalability reduce the reliability of intrusion detection in dynamic cloud infrastructures.</div></div><div><h3>Proposed model</h3><div>To address these, the paper proposes a Blockchain-Integrated IDS that contains an optimized Cosine Convolutional Neural Network enhanced with the Energy Valley Optimization Algorithm, CCNN-EVOA. The framework is designed for accurate detection with scalability and privacy-preserving data handling in real-time cloud environments.</div></div><div><h3>Key methods</h3><div>It integrates several components to achieve better performance of the IDS. A JDAGEM is used to carry out advanced pre-processing and feature extraction. Sensitive information is kept secure by PEDS. The Adaptive Blockchain Sharding Protocol allows for scalable, distributed, and tamper-resistant processing. The CCNN parameters are tuned using the EVOA algorithm for enhanced generalization and reduced false alarms.</div></div><div><h3>Results</h3><div>The framework is validated on real-world IoT-based attack scenarios using the BoT-IoT and UNSW-NB15 datasets. The proposed system has reached a high accuracy of 99.7% with notable improvements compared to state-of-the-art IDS techniques.</div></div><div><h3>Implications</h3><div>These findings affirm that Blockchain-Integrated IDS based on CCNN-EVOA provides a reliable, scalable, and privacy-preserving solution for contemporary IoT-enabled cloud infrastructures.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"735 ","pages":"Article 123015"},"PeriodicalIF":6.8,"publicationDate":"2025-12-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145978798","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-22DOI: 10.1016/j.ins.2025.123019
Madjid Tavana , Hosein Arman , Andreas Dellnitz
A variety of point-based heuristics and metaheuristics have been developed to approximate the global optimum of univariate functions. However, these Point-Based Search (PBS) algorithms often converge to local optima and fail to detect all extrema due to limited domain exploration. This study proposes an Area-Based Search (ABS) algorithm that systematically partitions the domain into uniformly spaced subintervals and evaluates the area under the curve in each segment. Subintervals with significantly larger or smaller areas than their neighbors are likely to contain local maxima or minima, respectively. We validate this idea on multimodal test functions using a Monte Carlo simulation framework with 1,000 trials. Across all noise levels in a standard benchmark function, ABS consistently detects all 16 local and global extrema. Intuitively, coverage measures the fraction of true extrema that an algorithm successfully recovers within a prescribed positional tolerance. Compared to Genetic Algorithms (GA), ABS achieves up to 37% higher detection accuracy under noise, with an average coverage improvement of 4.75% across all test cases. Additionally, ABS exhibited a 30.41% lower position error and a 36.89% lower value error than GA. The deterministic nature of ABS, with only one tunable resolution parameter, supports its use in noisy environments requiring full-spectrum extrema detection.
各种基于点的启发式方法和元启发式方法已经被发展用来逼近单变量函数的全局最优。然而，这些基于点的搜索（PBS）算法往往收敛于局部最优，并且由于有限的域探索而无法检测到所有的极值。本文提出了一种基于区域的搜索（area - based Search， ABS）算法，该算法系统地将区域划分为均匀间隔的子区间，并评估每个分段曲线下的面积。区域明显大于或小于相邻区域的子区间可能分别包含局部最大值或最小值。我们使用具有1000次试验的蒙特卡罗模拟框架在多模态测试函数上验证了这一想法。在标准基准功能的所有噪声级别中，ABS始终检测所有16个局部和全局极值。直观地说，覆盖率衡量的是算法在规定的位置容限内成功恢复的真极值的比例。与遗传算法（GA）相比，ABS在噪声下的检测精度提高了37%，所有测试用例的平均覆盖率提高了4.75%。ABS的位置误差比GA低30.41%，值误差比GA低36.89%。ABS的确定性特性，只有一个可调的分辨率参数，支持其在需要全频谱极端检测的嘈杂环境中使用。
{"title":"An area-based computational algorithm for robust extrema detection in noisy environments","authors":"Madjid Tavana , Hosein Arman , Andreas Dellnitz","doi":"10.1016/j.ins.2025.123019","DOIUrl":"10.1016/j.ins.2025.123019","url":null,"abstract":"<div><div>A variety of point-based heuristics and metaheuristics have been developed to approximate the global optimum of univariate functions. However, these Point-Based Search (PBS) algorithms often converge to local optima and fail to detect all extrema due to limited domain exploration. This study proposes an Area-Based Search (ABS) algorithm that systematically partitions the domain into uniformly spaced subintervals and evaluates the area under the curve in each segment. Subintervals with significantly larger or smaller areas than their neighbors are likely to contain local maxima or minima, respectively. We validate this idea on multimodal test functions using a Monte Carlo simulation framework with 1,000 trials. Across all noise levels in a standard benchmark function, ABS consistently detects all 16 local and global extrema. Intuitively, coverage measures the fraction of true extrema that an algorithm successfully recovers within a prescribed positional tolerance. Compared to Genetic Algorithms (GA), ABS achieves up to 37% higher detection accuracy under noise, with an average coverage improvement of 4.75% across all test cases. Additionally, ABS exhibited a 30.41% lower position error and a 36.89% lower value error than GA. The deterministic nature of ABS, with only one tunable resolution parameter, supports its use in noisy environments requiring full-spectrum extrema detection.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"734 ","pages":"Article 123019"},"PeriodicalIF":6.8,"publicationDate":"2025-12-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145841765","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-20DOI: 10.1016/j.ins.2025.123018
Peijian Zeng , Xingming Liao , Jianhui Xu , Shuisen Chen , Zhuowei Wang , Aimin Yang , Xingda Chen
Dissolved oxygen (DO) is a critical parameter for maintaining the ecological integrity of estuarine ecosystems. However, accurate DO prediction is hindered by measurement noise and complex periodic dynamics driven by tidal and seasonal cycles. To address these challenges, this study proposes a novel Recurrent Weighted Key-Value with State-Space Kalman and Fourier Filtering (RWKV-SKF) framework for enhanced DO forecasting. The RWKV-SKF integrates four specialized components: the Kalman Filtering Module (KFM) mitigates sensor noise; the Fourier Derivative Module (FDM) extracts dominant periodic features through spectral analysis; the Time Mix Module (TMM) captures short-term temporal dependencies; and the Channel Mix Module (CMM) models inter-variable interactions. By extending the RWKV architecture, the framework synergistically combines denoising, periodicity identification, and sequential learning. Experimental evaluations on both 4-hourly and daily DO monitoring datasets demonstrate that RWKV-SKF achieves state-of-the-art performance, reducing prediction errors by 0.97 % and 7.63 %, respectively, compared to the second-best model and attaining the lowest MSE of 0.5285 and 0.5499 among 19 baselines. These results highlight RWKV-SKF’s superior ability to handle noisy, cyclic DO dynamics, offering a robust solution for early warning and management of estuarine hypoxia.
{"title":"RWKV-SKF: A recurrent architecture with state-space and frequency-domain filtering for dissolved oxygen predicting and revealing influencing mechanisms","authors":"Peijian Zeng , Xingming Liao , Jianhui Xu , Shuisen Chen , Zhuowei Wang , Aimin Yang , Xingda Chen","doi":"10.1016/j.ins.2025.123018","DOIUrl":"10.1016/j.ins.2025.123018","url":null,"abstract":"<div><div>Dissolved oxygen (DO) is a critical parameter for maintaining the ecological integrity of estuarine ecosystems. However, accurate DO prediction is hindered by measurement noise and complex periodic dynamics driven by tidal and seasonal cycles. To address these challenges, this study proposes a novel Recurrent Weighted Key-Value with State-Space Kalman and Fourier Filtering (RWKV-SKF) framework for enhanced DO forecasting. The RWKV-SKF integrates four specialized components: the Kalman Filtering Module (KFM) mitigates sensor noise; the Fourier Derivative Module (FDM) extracts dominant periodic features through spectral analysis; the Time Mix Module (TMM) captures short-term temporal dependencies; and the Channel Mix Module (CMM) models inter-variable interactions. By extending the RWKV architecture, the framework synergistically combines denoising, periodicity identification, and sequential learning. Experimental evaluations on both 4-hourly and daily DO monitoring datasets demonstrate that RWKV-SKF achieves state-of-the-art performance, reducing prediction errors by 0.97 % and 7.63 %, respectively, compared to the second-best model and attaining the lowest MSE of 0.5285 and 0.5499 among 19 baselines. These results highlight RWKV-SKF’s superior ability to handle noisy, cyclic DO dynamics, offering a robust solution for early warning and management of estuarine hypoxia.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"734 ","pages":"Article 123018"},"PeriodicalIF":6.8,"publicationDate":"2025-12-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145841766","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-20DOI: 10.1016/j.ins.2025.123022
Conglin Li , Qingke Zhang , Junqing Li , Sichen Tao , Diego Oliva
Metaheuristic optimization algorithms have demonstrated strong performance when applied to complex nonlinear optimization tasks. However, their performance often degrades in high-dimensional and multimodal settings due to premature convergence and insufficient global search. To address these limitations, an Adversarial Game Optimization Algorithm (AGOA) is proposed, which constructs a metaheuristic optimization framework based on adversarial game mechanisms. AGOA integrates three mechanisms: (i) a dynamic role-based population partitioning strategy that assigns individuals as elites, explorers, or responders to balance exploration and exploitation; (ii) an adversarial feedback mechanism where worst-case responders introduce directed perturbations to counter elite dominance; and (iii) a diversity-preserving breakout strategy that monitors population stagnation and activates adaptive restarts. AGOA was tested on CEC 2017 (30D/50D/100D) and CEC 2022 (10D/20D), as well as applications including multi-threshold image segmentation, constrained engineering design, and UAV 3D path planning. Experimental evaluations indicate that AGOA achieves superior performance compared with 79 optimizers in solution quality, convergence behavior, and stability, achieving top rankings across all test categories. Theoretical analysis further establishes convergence to the global optimum in expectation and probability under mild conditions. Overall, AGOA offers a scalable and generalizable optimization framework with strong practical relevance. An open-access implementation of AGOA is provided at https://github.com/tsingke/AGOA.
{"title":"Adversarial game optimization: A game-theoretic metaheuristic for efficient complex optimization and engineering applications","authors":"Conglin Li , Qingke Zhang , Junqing Li , Sichen Tao , Diego Oliva","doi":"10.1016/j.ins.2025.123022","DOIUrl":"10.1016/j.ins.2025.123022","url":null,"abstract":"<div><div>Metaheuristic optimization algorithms have demonstrated strong performance when applied to complex nonlinear optimization tasks. However, their performance often degrades in high-dimensional and multimodal settings due to premature convergence and insufficient global search. To address these limitations, an Adversarial Game Optimization Algorithm (AGOA) is proposed, which constructs a metaheuristic optimization framework based on adversarial game mechanisms. AGOA integrates three mechanisms: (i) a dynamic role-based population partitioning strategy that assigns individuals as elites, explorers, or responders to balance exploration and exploitation; (ii) an adversarial feedback mechanism where worst-case responders introduce directed perturbations to counter elite dominance; and (iii) a diversity-preserving breakout strategy that monitors population stagnation and activates adaptive restarts. AGOA was tested on CEC 2017 (30D/50D/100D) and CEC 2022 (10D/20D), as well as applications including multi-threshold image segmentation, constrained engineering design, and UAV 3D path planning. Experimental evaluations indicate that AGOA achieves superior performance compared with 79 optimizers in solution quality, convergence behavior, and stability, achieving top rankings across all test categories. Theoretical analysis further establishes convergence to the global optimum in expectation and probability under mild conditions. Overall, AGOA offers a scalable and generalizable optimization framework with strong practical relevance. An open-access implementation of AGOA is provided at <span><span>https://github.com/tsingke/AGOA</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"735 ","pages":"Article 123022"},"PeriodicalIF":6.8,"publicationDate":"2025-12-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145852495","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-19DOI: 10.1016/j.ins.2025.122996
Peng Wang , Yan Wang , Jie Jin
Nonlinear problems are frequently encountered in engineering fields, and most of the nonlinear problems can be modeled as nonlinear equations (NEs). Accordingly, quickly and correctly obtaining the solutions of NEs becomes particularly critical. However, most NEs have multiple roots, and efficiently locating all roots remains a challenging task. To address this issue, this work proposes a density-based spatial clustering of applications with noise (DBSCAN)-enhanced niche differential evolution (DBNDE) algorithm for solving NEs. By integrating a density-based clustering technique with niching technology, the proposed DBNDE algorithm achieves effective coordination between global exploration and local refinement. Specifically, the DBNDE algorithm dynamically partitions the population using density clustering, categorizing individuals into noise points, suboptimal solution clusters, and optimal solution clusters, thereby enhancing the efficiency of root identification. Additionally, it incorporates migration strategies to optimize solution distribution and employs an archive mechanism to preserve optimal solutions, improving the quality and stability of results. To substantiate the superiority of DBNDE, we carried out head-to-head comparisons with several classical algorithms on a benchmark set comprising thirty widely used and ten newly added NEs. Then, we conducted performance evaluations on two high-dimensional nonlinear problems. Results demonstrate that the proposed DBNDE algorithm effectively locates multiple roots of the NEs in a single run. Furthermore, the core parameter sensitivity analysis of the proposed DBNDE algorithm reveals its strong robustness to parameter settings, ensuring its stable performance across diverse problem sets.
{"title":"A DBSCAN-enhanced niching differential evolution algorithm for solving nonlinear equations","authors":"Peng Wang , Yan Wang , Jie Jin","doi":"10.1016/j.ins.2025.122996","DOIUrl":"10.1016/j.ins.2025.122996","url":null,"abstract":"<div><div>Nonlinear problems are frequently encountered in engineering fields, and most of the nonlinear problems can be modeled as nonlinear equations (NEs). Accordingly, quickly and correctly obtaining the solutions of NEs becomes particularly critical. However, most NEs have multiple roots, and efficiently locating all roots remains a challenging task. To address this issue, this work proposes a density-based spatial clustering of applications with noise (DBSCAN)-enhanced niche differential evolution (DBNDE) algorithm for solving NEs. By integrating a density-based clustering technique with niching technology, the proposed DBNDE algorithm achieves effective coordination between global exploration and local refinement. Specifically, the DBNDE algorithm dynamically partitions the population using density clustering, categorizing individuals into noise points, suboptimal solution clusters, and optimal solution clusters, thereby enhancing the efficiency of root identification. Additionally, it incorporates migration strategies to optimize solution distribution and employs an archive mechanism to preserve optimal solutions, improving the quality and stability of results. To substantiate the superiority of DBNDE, we carried out head-to-head comparisons with several classical algorithms on a benchmark set comprising thirty widely used and ten newly added NEs. Then, we conducted performance evaluations on two high-dimensional nonlinear problems. Results demonstrate that the proposed DBNDE algorithm effectively locates multiple roots of the NEs in a single run. Furthermore, the core parameter sensitivity analysis of the proposed DBNDE algorithm reveals its strong robustness to parameter settings, ensuring its stable performance across diverse problem sets.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"734 ","pages":"Article 122996"},"PeriodicalIF":6.8,"publicationDate":"2025-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145841772","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号