Information Sciences最新文献

Global exponential synchronization of competitive neural networks with D operators and mixed delays and an application to secure communication 具有D算子和混合延迟的竞争神经网络的全局指数同步及其在安全通信中的应用

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-07-02 DOI: 10.1016/j.ins.2025.122473

Weijing Yan , Yu Xue , Xian Zhang

引用次数: 0

Multi-objective CNN optimization: A robust framework for automated model design 多目标CNN优化：自动化模型设计的鲁棒框架

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-07-01 DOI: 10.1016/j.ins.2025.122468

Sefa Aras , Elif Aras , Eyüp Gedikli , Hamdi Tolga Kahraman

{"title":"Multi-objective CNN optimization: A robust framework for automated model design","authors":"Sefa Aras , Elif Aras , Eyüp Gedikli , Hamdi Tolga Kahraman","doi":"10.1016/j.ins.2025.122468","DOIUrl":"10.1016/j.ins.2025.122468","url":null,"abstract":"<div><div>Convolutional neural networks (CNNs) have demonstrated high performance in classifying image data. However, CNNs require expert-level hyperparameter tuning and involve substantial computational complexity, which hinders their effective deployment in real-time and IoT systems. Research on CNNs indicates that optimal hyperparameter configurations can enhance inference speed while improving classification accuracy. We developed a novel approach that uses multi-objective optimization to design CNN models automatically. Our method tunes hyperparameters to balance classification accuracy and inference speed. We define classification performance and inference speed as objectives and balance them using a Pareto-optimal strategy. Unlike traditional approaches, MoCNN systematically explores Pareto-optimal trade-offs between classification performance and computational efficiency, enabling a fully automated architecture search without manual intervention. In this study, we select NSGA-II as our preferred MOEA while ensuring the framework remains flexible enough to accommodate other evolutionary strategies. Experimental evaluations on benchmark datasets (CIFAR-10, CIFAR-100, and FRUITS-360) demonstrate that MoCNN reduces inference time by up to 72.02% on average and improves classification accuracy by 6.72% compared to manually tuned CNN architectures. By eliminating the need for heuristic hyperparameter selection, MoCNN enhances scalability and is particularly well suited for real-time, mobile AI, and edge-computing applications. Our results show that MoCNN outperforms state-of-the-art optimization frameworks in both computational efficiency and predictive performance, highlighting its potential for deployment in scenarios where accuracy and speed are critical.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"719 ","pages":"Article 122468"},"PeriodicalIF":8.1,"publicationDate":"2025-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144524234","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}

引用次数: 0

DEEP-CWS: Distilling Efficient pre-trained models with Early exit and Pruning for scalable Chinese Word Segmentation DEEP-CWS：基于早期退出和修剪的高效预训练模型的可扩展中文分词

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-07-01 DOI: 10.1016/j.ins.2025.122470

Shiting Xu

引用次数: 0

A coordinator-driven consensus-reaching model for green technology utilisation rate determination 绿色技术利用率确定的协调者驱动共识达成模型

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-06-28 DOI: 10.1016/j.ins.2025.122472

Nana Liu , Xianzhe Zhang , Hangyao Wu , Bai Yang , Yuelong Zheng

{"title":"A coordinator-driven consensus-reaching model for green technology utilisation rate determination","authors":"Nana Liu , Xianzhe Zhang , Hangyao Wu , Bai Yang , Yuelong Zheng","doi":"10.1016/j.ins.2025.122472","DOIUrl":"10.1016/j.ins.2025.122472","url":null,"abstract":"<div><div>Determining the green technology utilisation rate is the initial process of green building development. However, due to the different interest demands of stakeholders, it’s difficult to reach a consensus on the determination of the green technology utilisation rate. The social network-based consensus-reaching process (CRP) can help mitigate stakeholder discrepancies. As a result, we determine the green technology utilisation rate through the social network-based CRP. However, existing social network-based CRP methods assign weights to stakeholders from a single perspective, failing to meet the objectives of all involved parties. Additionally, these methods cannot guide the direction of opinion change, making them unsuitable for determining the green technology utilisation rate. Moreover, the cost measure during the CRP is often oversimplified to a single unit cost, which lacks precision. To address these limitations, we develop a coordinator-driven consensus-reaching model. In our model, stakeholders’ weights are assigned using two distinct methods, and a unique adjustment mechanism is designed to gradually guide opinions towards the desired outcome. We also separately measure the financial cost and time cost to obtain a more accurate result. Finally, a numerical example and several simulation experiments are conducted to demonstrate the effectiveness of our model.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"719 ","pages":"Article 122472"},"PeriodicalIF":8.1,"publicationDate":"2025-06-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144517281","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}

引用次数: 0

A deep reinforcement learning-based controller design framework for Lipschitz continuous nonlinear systems 基于深度强化学习的Lipschitz连续非线性系统控制器设计框架

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-06-27 DOI: 10.1016/j.ins.2025.122455

Yuan Li , Siyang Zhao , Jinyong Yu

{"title":"A deep reinforcement learning-based controller design framework for Lipschitz continuous nonlinear systems","authors":"Yuan Li , Siyang Zhao , Jinyong Yu","doi":"10.1016/j.ins.2025.122455","DOIUrl":"10.1016/j.ins.2025.122455","url":null,"abstract":"<div><div>Due to the complex dynamics and uncertainty of the nonlinear systems, designing controllers for such systems poses significant challenges. To address this dilemma, deep reinforcement learning (DRL) indicates a promising method. However, most designs of reward/value functions in DRL rely on experience, which takes much trial and error. In order to decrease the trial cost, this paper proposes a novel DRL method founded on actor-critic (AC) architecture for nonlinear system controller design, which is called actor-Lyapunov (AL). Diverging from conventional AC architecture, AL eliminates the necessity of the critic network. The actor network can be trained by utilizing a kind of Lyapunov function as the value function. Firstly, we provide a perspective of normed linear space to clarify the controller design. The controller generated by the actor network is regarded as a proper mapping within the state space. Based on this concept, the convergence of this approach under gradient descent is briefly analyzed. Next, a refined value function related to the exponent is introduced to promote the training effect of the actor network. Finally, simulations are conducted to validate the efficacy of our approach and illustrate the advantages of the refined value function in improving system performance.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"719 ","pages":"Article 122455"},"PeriodicalIF":8.1,"publicationDate":"2025-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144517282","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}

引用次数: 0

Efficient algorithms for parameter-free edge structural diversity search on graphs 图上无参数边结构多样性搜索的高效算法

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-06-27 DOI: 10.1016/j.ins.2025.122454

Yuting Tan, Junfeng Zhou, Ming Du

{"title":"Efficient algorithms for parameter-free edge structural diversity search on graphs","authors":"Yuting Tan, Junfeng Zhou, Ming Du","doi":"10.1016/j.ins.2025.122454","DOIUrl":"10.1016/j.ins.2025.122454","url":null,"abstract":"<div><div>Edge structural diversity refers to the complexity of the social context in the common neighbors of a pair of vertices, which can be used as an important indicator of the spread of information and social influence. The existing edge structural diversity score is designed based on the given threshold parameter, which is easy to vary with the parameter, leading to unstable results. In this paper, we propose an edge diversity model without the threshold parameter to get stable results, based on which we propose two query problems, top-<em>k</em> and skyline edge search. We propose the upper-bound online algorithms, which obtain results by computing exact diversity scores for partial edges. Then, we propose the basic index. Based on this basic index we can obtain edges of each ego-network directly, without extracting the ego-network from the original graph. Further, we propose the optimized index, which maps edges in the original graph as super-vertices and records information using super-edges, reducing the index size. Finally, we conduct experiments on 12 real-world datasets. The experimental results verify the effectiveness and efficiency of our algorithms.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"719 ","pages":"Article 122454"},"PeriodicalIF":8.1,"publicationDate":"2025-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144517284","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}

引用次数: 0

Double event-triggered based anti-disturbance optimal control for nonlinear systems using adaptive dynamic programming 基于双事件触发的非线性系统自适应动态规划抗干扰最优控制

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-06-26 DOI: 10.1016/j.ins.2025.122453

Wenyang Su , Yang Yi , Mouquan Shen , Guangyu Zhu , Songyin Cao

{"title":"Double event-triggered based anti-disturbance optimal control for nonlinear systems using adaptive dynamic programming","authors":"Wenyang Su , Yang Yi , Mouquan Shen , Guangyu Zhu , Songyin Cao","doi":"10.1016/j.ins.2025.122453","DOIUrl":"10.1016/j.ins.2025.122453","url":null,"abstract":"<div><div>This article presents a robust anti-disturbance optimal control approach for nonlinear systems with different disturbances, utilizing adaptive dynamic programming (ADP) techniques. Through the integration of zero-sum (ZS) game and nonlinear disturbance observer (NDO), we not only address the challenge of controlling multiple disturbances, but also achieve precise anti-disturbance results. The problem of dealing with mismatched disturbances is formulated as a ZS game, and then the optimal control strategy under the worst-case disturbances is derived within the framework of ADP. Meanwhile, the NDO is specifically tailored to estimate the dynamics of those disturbances occurring at input port. Furthermore, a novel double event-triggered mechanism has been developed to alleviate the burden of actuators resulting from state transfer and disturbance estimation. Its unique feature is based on the principle of first-arrival and same-trigger while ensuring synchronization of signal transmission. Depending on the estimated disturbances and the designed weight update law with critic neural network, a control input is generated to guarantee that the closed-loop systems are asymptotically stable. Numerous simulations have been conducted to validate the efficacy of the proposed algorithm in terms of conserving bandwidth resources and anti-disturbance control. The suitability for practical application in autonomous underwater vehicles (AUVs) has further been affirmed through the Gazebo AUV simulator.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"719 ","pages":"Article 122453"},"PeriodicalIF":8.1,"publicationDate":"2025-06-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144489872","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}

引用次数: 0

Feature selection optimization algorithm based on evolutionary Q-learning 基于进化q学习的特征选择优化算法

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-06-26 DOI: 10.1016/j.ins.2025.122441

Guan Yang, Zhiyong Zeng, Xinrui Pu, Ren Duan

{"title":"Feature selection optimization algorithm based on evolutionary Q-learning","authors":"Guan Yang, Zhiyong Zeng, Xinrui Pu, Ren Duan","doi":"10.1016/j.ins.2025.122441","DOIUrl":"10.1016/j.ins.2025.122441","url":null,"abstract":"<div><div>Classification problems are an important research area in the field of data mining and machine learning. To enhance classification accuracy and optimize the effectiveness of learning algorithms, feature selection, as a data preprocessing operation, deserves ongoing attention. Based on reinforcement learning and particle swarm optimization, this paper proposes an evolutionary Q-learning feature selection optimization algorithm (EQL-FS). It leverages the advantages of reinforcement learning and combines them with the global exploration capability of the particle swarm optimization algorithm to achieve the optimal strategy. The multiagent approach is adopted, and the interaction is achieved through particle swarm optimization. The effectiveness of the proposed algorithm has been validated using sixteen public datasets. The experimental results indicate that this new algorithm can select the shortest feature subset without compromising accuracy. Additionally, it demonstrates a robust ability to eliminate noise and redundant features. Furthermore, the algorithm has been applied to analyze the broadband customer base churn for a communication operator, and the results are consistent with those obtained from the public datasets. Finally, the statistical test comparing different algorithms has been completed, and the results indicate that the new algorithm EQL-FS demonstrates statistical significance in terms of accuracy and the number of selected features.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"719 ","pages":"Article 122441"},"PeriodicalIF":8.1,"publicationDate":"2025-06-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144490000","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}

引用次数: 0

Dynamic sequential neighbor processing: A liquid neural network-inspired framework for enhanced graph neural networks 动态顺序邻居处理：一种基于液体神经网络的增强图神经网络框架

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-06-26 DOI: 10.1016/j.ins.2025.122452

Kuijie Zhang , Shanchen Pang , Yuanyuan Zhang , Yun Bai , Luqi Wang , Jerry Chun-Wei Lin

{"title":"Dynamic sequential neighbor processing: A liquid neural network-inspired framework for enhanced graph neural networks","authors":"Kuijie Zhang , Shanchen Pang , Yuanyuan Zhang , Yun Bai , Luqi Wang , Jerry Chun-Wei Lin","doi":"10.1016/j.ins.2025.122452","DOIUrl":"10.1016/j.ins.2025.122452","url":null,"abstract":"<div><div>Integrating information from multi-order neighborhoods is a fundamental strategy in Graph Neural Networks (GNNs) for capturing higher-order structural patterns and enhancing the expressive power of node representations. However, most existing GNNs treat neighbors from different orders as unordered sets and integrate them using static or parallel strategies, thus overlooking the sequential and evolving nature of neighborhood expansion. To address this limitation, we propose a novel GNN framework, SL, which integrates <strong>Serialized Neighbor Features</strong> with <strong>Liquid Neural Networks</strong> (LNNs) to enable order-aware, dynamic adaptation of neighbor influence. By modeling neighbor features as ordered sequences and leveraging LNNs' internal feedback dynamics, SL adapts feature extraction in real time based on local context and propagation history. This design offers fine-grained control over hierarchical dependencies and allows dynamic modulation of contributions from different neighborhood layers. SL is model-agnostic and can be seamlessly integrated with both classical and state-of-the-art GNNs. Extensive experiments across ten benchmark datasets show that SL consistently improves node classification accuracy and significantly alleviates over-smoothing in deep GNNs. These results highlight that order-aware and dynamically regulated propagation represents a powerful, flexible alternative to traditional multi-order aggregation, enhancing the adaptability and expressiveness of GNNs for complex graph learning tasks.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"719 ","pages":"Article 122452"},"PeriodicalIF":8.1,"publicationDate":"2025-06-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144489999","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}

引用次数: 0

TeProM: A rule-free method for extracting process from complex text with enhanced coreference handling TeProM：一种无规则的方法，用于从具有增强的共同引用处理的复杂文本中提取过程

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-06-26 DOI: 10.1016/j.ins.2025.122451

Xiaoxiao Sun , Chenying Zhao , Dongjin Yu , Yi Xu , Nana Xiao

{"title":"TeProM: A rule-free method for extracting process from complex text with enhanced coreference handling","authors":"Xiaoxiao Sun , Chenying Zhao , Dongjin Yu , Yi Xu , Nana Xiao","doi":"10.1016/j.ins.2025.122451","DOIUrl":"10.1016/j.ins.2025.122451","url":null,"abstract":"<div><div>Extracting business process models from textual documents remains a significant challenge in enterprises. Traditional rule-based methods suffer from poor applicability due to customized rule sets while most machine-learning based methods focus on simple process documents. This paper presents Text-based Process Modeling (TeProM), a novel method for extracting business process components and their relations from textual descriptions. By adopting a rule-free design, TeProM departs from traditional rule-based systems and leverages a neural network model to address complex coreference phenomena in text, thereby ensuring the accurate mapping of process components within the model. This approach applies to various types of business process documents, particularly excelling in processing complex textual structures with long-range dependencies. Compared to previous approaches, TeProM is able to effectively address the complex logical structures and coreference issues concealed in business process documents. TeProM achieved the best performance over 10 baselines in multidimensional evaluation. Additionally, evaluations on the PET and SAP-OPC datasets for relation extraction further demonstrated the effectiveness of the proposed method. An annotated dataset consisting of 91 real business process documents is also provided, which serves as a valuable resource for future research.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"719 ","pages":"Article 122451"},"PeriodicalIF":8.1,"publicationDate":"2025-06-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144517283","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}

引用次数: 0