Information Sciences最新文献_第5页

Reinforcement learning applied to a situation awareness decision-making model

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2025-02-07 DOI: 10.1016/j.ins.2025.121928

Renato D. Costa, Celso M. Hirata

Situation awareness is critical for successful decision-making in safety–critical and mission-critical environments such as air traffic and electric power control rooms. Situation awareness models provide high explainability; however, the decision support systems based on these models require the intervention of experts for initial configuration and evolutionary maintenance tasks, which are generally costly. Reinforcement learning is a machine learning strategy that considers how software agents act in an environment to maximize some cumulative reward by improving performance through experience. We investigated how reinforcement learning can help experts configure and maintain situation awareness models. This work proposes the Reinforcement Learning Situation Awareness (RLSA) method to automate the initial and evolving set-ups of the cognitive model’s belief parameters of situation awareness models employed by decision support systems using reinforcement learning techniques. Tests applying the method on a simulated case study and public datasets with distinct evolving and non-evolving conditions, using accuracy and other metrics, show promising results compared to those found in literature, including baseline Naïve Bayes and Decision Tree algorithms. The effectiveness in automating the parameter adjustments shown by RLSA reduces the demand for specialized work in applications with evolving behavior while maintaining the explainable cognitive characteristics of situation awareness models.

{"title":"Reinforcement learning applied to a situation awareness decision-making model","authors":"Renato D. Costa, Celso M. Hirata","doi":"10.1016/j.ins.2025.121928","DOIUrl":"10.1016/j.ins.2025.121928","url":null,"abstract":"<div><div>Situation awareness is critical for successful decision-making in safety–critical and mission-critical environments such as air traffic and electric power control rooms. Situation awareness models provide high explainability; however, the decision support systems based on these models require the intervention of experts for initial configuration and evolutionary maintenance tasks, which are generally costly. Reinforcement learning is a machine learning strategy that considers how software agents act in an environment to maximize some cumulative reward by improving performance through experience. We investigated how reinforcement learning can help experts configure and maintain situation awareness models. This work proposes the Reinforcement Learning Situation Awareness (RLSA) method to automate the initial and evolving set-ups of the cognitive model’s belief parameters of situation awareness models employed by decision support systems using reinforcement learning techniques. Tests applying the method on a simulated case study and public datasets with distinct evolving and non-evolving conditions, using accuracy and other metrics, show promising results compared to those found in literature, including baseline Naïve Bayes and Decision Tree algorithms. The effectiveness in automating the parameter adjustments shown by RLSA reduces the demand for specialized work in applications with evolving behavior while maintaining the explainable cognitive characteristics of situation awareness models.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"704 ","pages":"Article 121928"},"PeriodicalIF":8.1,"publicationDate":"2025-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143422759","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

Adaptive mutation based on multi-population evolution strategy for greybox fuzzing

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2025-02-07 DOI: 10.1016/j.ins.2025.121959

Weihua Jiao , Xilong Li , Qingbao Li , Fei Cao , Xiaonan Li , Shudan Yue

The development of adaptive mutation techniques to enhance gray-box fuzzing performance has become trendy. However, existing adaptive methods have limitations in that they either ignore the impacts of different characteristics of seed inputs or require assumptions about the probability distribution model. Motivated by the observation, we present a novel adaptive mutation approach that combines seed clustering and Evolution Strategy to automatically find the optimal mutation scheduling method for seeds with different characteristics. Our approach captures seed inputs' structural and functional similarities and partitions them into proper populations. The Evolution Strategy is then used to iteratively optimize the probability distribution of operator selection for each population. We implement the prototype tool MesFuzz based on the aforementioned ideas. Evaluation on LAVA-M shows that MesFuzz is the only fuzzer to find bugs in all target programs. In addition, MesFuzz improves the path coverage by 132%, 14%, and 12% over DARWIN, SeamFuzz, and AFL++, respectively. That will facilitate fuzzing to discover vulnerabilities in real-world software and firmware further.

{"title":"Adaptive mutation based on multi-population evolution strategy for greybox fuzzing","authors":"Weihua Jiao , Xilong Li , Qingbao Li , Fei Cao , Xiaonan Li , Shudan Yue","doi":"10.1016/j.ins.2025.121959","DOIUrl":"10.1016/j.ins.2025.121959","url":null,"abstract":"<div><div>The development of adaptive mutation techniques to enhance gray-box fuzzing performance has become trendy. However, existing adaptive methods have limitations in that they either ignore the impacts of different characteristics of seed inputs or require assumptions about the probability distribution model. Motivated by the observation, we present a novel adaptive mutation approach that combines seed clustering and Evolution Strategy to automatically find the optimal mutation scheduling method for seeds with different characteristics. Our approach captures seed inputs' structural and functional similarities and partitions them into proper populations. The Evolution Strategy is then used to iteratively optimize the probability distribution of operator selection for each population. We implement the prototype tool MesFuzz based on the aforementioned ideas. Evaluation on LAVA-M shows that <span>MesFuzz</span> is the only fuzzer to find bugs in all target programs. In addition, <span>MesFuzz</span> improves the path coverage by 132%, 14%, and 12% over DARWIN, <span>SeamFuzz</span>, and AFL++, respectively. That will facilitate fuzzing to discover vulnerabilities in real-world software and firmware further.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"705 ","pages":"Article 121959"},"PeriodicalIF":8.1,"publicationDate":"2025-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143428695","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

A2SHE: An anonymous authentication scheme for health emergencies in public venues

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2025-02-07 DOI: 10.1016/j.ins.2025.121944

Xiaohan Yue , Peng Yang , Haoran Si , Haibo Yang , Fucai Zhou , Qiang Wang , Zhuo Yang , Shi Bai , Yuan He

With the outbreak of health emergencies such as COVID-19 and monkeypox, individuals are required to present their pandemic prevention and control (PPC) credentials when accessing public venues, which raises significant privacy concerns that the data from individuals embedded in the PPC credentials may be leaked maliciously. While previous studies have proposed privacy-preserving solutions, they come with their own set of challenges. Specifically, none can simultaneously meet the requirements of non-transferability for the information in authentication and privacy protection for personal identities. Moreover, the existing membership management mechanisms adopted by revocable solutions entail additional computational costs for both users and verifiers. Therefore, this paper presents an anonymous authentication scheme for health emergencies in public venues (A²SHE, for short) to overcome these challenges. A²SHE offers a novel membership management mechanism that supports revocability for invalid users and traceability for patients while considering the trade-off between privacy and performance. A²SHE also introduces a biometric-based key derivation algorithm to prevent the transferability of authentication information. Furthermore, based on the framework of A²SHE and considering the security and privacy requirements for health emergencies in public venues, a concrete construction of A²SHE is presented. The feasibility of the proposed scheme is demonstrated through both of the security and performance analysis. The results show that A²SHE achieves an optimal balance between security and performance that differs from previous schemes, presenting a novel practical approach for access control in public venues.

{"title":"A2SHE: An anonymous authentication scheme for health emergencies in public venues","authors":"Xiaohan Yue , Peng Yang , Haoran Si , Haibo Yang , Fucai Zhou , Qiang Wang , Zhuo Yang , Shi Bai , Yuan He","doi":"10.1016/j.ins.2025.121944","DOIUrl":"10.1016/j.ins.2025.121944","url":null,"abstract":"<div><div>With the outbreak of health emergencies such as COVID-19 and monkeypox, individuals are required to present their pandemic prevention and control (PPC) credentials when accessing public venues, which raises significant privacy concerns that the data from individuals embedded in the PPC credentials may be leaked maliciously. While previous studies have proposed privacy-preserving solutions, they come with their own set of challenges. Specifically, none can simultaneously meet the requirements of non-transferability for the information in authentication and privacy protection for personal identities. Moreover, the existing membership management mechanisms adopted by revocable solutions entail additional computational costs for both users and verifiers. Therefore, this paper presents an anonymous authentication scheme for health emergencies in public venues (A<sup>2</sup>SHE, for short) to overcome these challenges. A<sup>2</sup>SHE offers a novel membership management mechanism that supports revocability for invalid users and traceability for patients while considering the trade-off between privacy and performance. A<sup>2</sup>SHE also introduces a biometric-based key derivation algorithm to prevent the transferability of authentication information. Furthermore, based on the framework of A<sup>2</sup>SHE and considering the security and privacy requirements for health emergencies in public venues, a concrete construction of A<sup>2</sup>SHE is presented. The feasibility of the proposed scheme is demonstrated through both of the security and performance analysis. The results show that A<sup>2</sup>SHE achieves an optimal balance between security and performance that differs from previous schemes, presenting a novel practical approach for access control in public venues.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"703 ","pages":"Article 121944"},"PeriodicalIF":8.1,"publicationDate":"2025-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143349382","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

NEEP-ADF: Neuro-encoded expression programming with automatically defined functions

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2025-02-07 DOI: 10.1016/j.ins.2025.121957

Jun Ma , Haoran Shan , Fengyang Sun , Lin Wang , Shuangrong Liu , Houguan Zhu , Fenghui Gao , Junteng Zheng , Bo Yang , Qinfei Li

Symbolic equations are crucial for scientific discovery. Symbolic regression, the task of extracting underlying mathematical expressions from data, represents a challenge in artificial intelligence. Although recent algorithms integrating symbolic regression with neural networks have emerged in the machine learning community, these approaches primarily focus on small-scale dependencies among symbols, neglecting relationships between larger scales (such as among substructures) and interactions between small and large scales (such as between symbols and substructures). A single-scale generation model can lead to redundant expression structures and convergence oscillations. This paper introduces Neuro-Encoded Expression Programming with Automatically Defined Functions (NEEP-ADF), a novel method addressing these challenges by learning multi-scale relationships. The NEEP-ADF method is based on two core ideas: 1) Symbols form reusable substructure modules through small-scale dependencies. 2) The model captures large-scale relationships among substructures to adapt to specific target problems. This multi-scale approach endows NEEP-ADF with flexible scalability, enabling it to dynamically adjust the complexity of solutions through symbols and substructures, thereby effectively addressing the problem of unknown scale. In a series of benchmark tests encompassing synthetic and real-world benchmarks, both versions of NEEP-ADF (Evolutionary Computation and Reinforcement Learning) demonstrated the state-of-the-art performance and convergence speed among the compared algorithms.

{"title":"NEEP-ADF: Neuro-encoded expression programming with automatically defined functions","authors":"Jun Ma , Haoran Shan , Fengyang Sun , Lin Wang , Shuangrong Liu , Houguan Zhu , Fenghui Gao , Junteng Zheng , Bo Yang , Qinfei Li","doi":"10.1016/j.ins.2025.121957","DOIUrl":"10.1016/j.ins.2025.121957","url":null,"abstract":"<div><div>Symbolic equations are crucial for scientific discovery. Symbolic regression, the task of extracting underlying mathematical expressions from data, represents a challenge in artificial intelligence. Although recent algorithms integrating symbolic regression with neural networks have emerged in the machine learning community, these approaches primarily focus on small-scale dependencies among symbols, neglecting relationships between larger scales (such as among substructures) and interactions between small and large scales (such as between symbols and substructures). A single-scale generation model can lead to redundant expression structures and convergence oscillations. This paper introduces Neuro-Encoded Expression Programming with Automatically Defined Functions (NEEP-ADF), a novel method addressing these challenges by learning multi-scale relationships. The NEEP-ADF method is based on two core ideas: 1) Symbols form reusable substructure modules through small-scale dependencies. 2) The model captures large-scale relationships among substructures to adapt to specific target problems. This multi-scale approach endows NEEP-ADF with flexible scalability, enabling it to dynamically adjust the complexity of solutions through symbols and substructures, thereby effectively addressing the problem of unknown scale. In a series of benchmark tests encompassing synthetic and real-world benchmarks, both versions of NEEP-ADF (Evolutionary Computation and Reinforcement Learning) demonstrated the state-of-the-art performance and convergence speed among the compared algorithms.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"704 ","pages":"Article 121957"},"PeriodicalIF":8.1,"publicationDate":"2025-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143377335","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

Enhancing efficiency and data utility in longitudinal data anonymization

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2025-02-06 DOI: 10.1016/j.ins.2025.121949

Fatemeh Amiri , David Sánchez , Josep Domingo-Ferrer

Longitudinal data consist of observations collected over time from a set of individuals. The accumulation of information on each individual over time makes longitudinal data particularly privacy-sensitive. However, existing anonymization methods are often inadequate for ensuring privacy-preserving publication of such data, as current privacy models assume unrealistic levels of attacker knowledge. To address this, we propose the

{(k, β)}^{L}

-privacy model, which assumes that an attacker's knowledge is limited to a subsequence of L quasi-identifiers. This provides a more realistic representation of the information an attacker might actually possess. Our model guarantees that every subsequence of L quasi-identifier values appears in either zero or at least k records within the longitudinal database. Additionally, it ensures that the confidence of any sensitive value within these k records is at most β times higher than its confidence in the entire dataset. This not only strengthens privacy protection but also enhances data utility.

Furthermore, we introduce FCLA, an anonymization algorithm designed to enforce our privacy model while prioritizing data utility. FCLA effectively mitigates identity and attribute disclosures, as well as skewness attacks in longitudinal data. It achieves this by partitioning sequences into groups and anonymizing them independently—a process that can be efficiently parallelized. Experimental results show that FCLA outperforms existing methods in preserving data utility while adhering to strict privacy constraints. Additionally, time complexity analysis and execution time measurements demonstrate that FCLA is more efficient and scalable than alternative approaches.

{"title":"Enhancing efficiency and data utility in longitudinal data anonymization","authors":"Fatemeh Amiri , David Sánchez , Josep Domingo-Ferrer","doi":"10.1016/j.ins.2025.121949","DOIUrl":"10.1016/j.ins.2025.121949","url":null,"abstract":"<div><div>Longitudinal data consist of observations collected over time from a set of individuals. The accumulation of information on each individual over time makes longitudinal data particularly privacy-sensitive. However, existing anonymization methods are often inadequate for ensuring privacy-preserving publication of such data, as current privacy models assume unrealistic levels of attacker knowledge. To address this, we propose the <span><math><msup><mrow><mo>(</mo><mi>k</mi><mo>,</mo><mi>β</mi><mo>)</mo></mrow><mrow><mi>L</mi></mrow></msup></math></span>-privacy model, which assumes that an attacker's knowledge is limited to a subsequence of <em>L</em> quasi-identifiers. This provides a more realistic representation of the information an attacker might actually possess. Our model guarantees that every subsequence of <em>L</em> quasi-identifier values appears in either zero or at least <em>k</em> records within the longitudinal database. Additionally, it ensures that the confidence of any sensitive value within these <em>k</em> records is at most <em>β</em> times higher than its confidence in the entire dataset. This not only strengthens privacy protection but also enhances data utility.</div><div>Furthermore, we introduce FCLA, an anonymization algorithm designed to enforce our privacy model while prioritizing data utility. FCLA effectively mitigates identity and attribute disclosures, as well as skewness attacks in longitudinal data. It achieves this by partitioning sequences into groups and anonymizing them independently—a process that can be efficiently parallelized. Experimental results show that FCLA outperforms existing methods in preserving data utility while adhering to strict privacy constraints. Additionally, time complexity analysis and execution time measurements demonstrate that FCLA is more efficient and scalable than alternative approaches.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"704 ","pages":"Article 121949"},"PeriodicalIF":8.1,"publicationDate":"2025-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143378820","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 fine-tuning of vision transformer via path-augmented parameter adaptation

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2025-02-06 DOI: 10.1016/j.ins.2025.121948

Yao Zhou , Zhang Yi , Gary G. Yen

Fine-tuning pre-trained Vision Transformer (ViT) models have been adopted as the de facto paradigm for achieving promising performance on visual tasks. However, the exponential growth in parameter size presents significant challenges to computational and storage efficiency when transferring ViT models to downstream tasks. Leveraging the assumption that trained models are over-parameterized and intrinsically reside a lower-dimensional space, learning a small number of parameters while freezing the backbone has emerged as a promising strategy for efficiently fine-tuning ViT models. In this paper, a path-augmented parameter adaptation method, termed as PPA, is proposed for fine-tuning ViT models. Specifically, a multi-path strategy is designed to learn the parameter updates in pre-trained ViT models, which aims to promote information flow and subspace representation learning via augmented paths. Based on this design, heterogeneous modules with a few learnable parameters are adopted which enable augmented paths to capture diverse information in low-dimensional subspaces. Since the parameters in the augmented paths can be reparametrized to the pre-trained model after fine-tuning, the proposed method incurs no additional inference cost. Extensive experiments and comparisons conducted on various visual benchmark tasks demonstrate the effectiveness of the proposed PPA method.

{"title":"Efficient fine-tuning of vision transformer via path-augmented parameter adaptation","authors":"Yao Zhou , Zhang Yi , Gary G. Yen","doi":"10.1016/j.ins.2025.121948","DOIUrl":"10.1016/j.ins.2025.121948","url":null,"abstract":"<div><div>Fine-tuning pre-trained Vision Transformer (ViT) models have been adopted as the de facto paradigm for achieving promising performance on visual tasks. However, the exponential growth in parameter size presents significant challenges to computational and storage efficiency when transferring ViT models to downstream tasks. Leveraging the assumption that trained models are over-parameterized and intrinsically reside a lower-dimensional space, learning a small number of parameters while freezing the backbone has emerged as a promising strategy for efficiently fine-tuning ViT models. In this paper, a path-augmented parameter adaptation method, termed as PPA, is proposed for fine-tuning ViT models. Specifically, a multi-path strategy is designed to learn the parameter updates in pre-trained ViT models, which aims to promote information flow and subspace representation learning via augmented paths. Based on this design, heterogeneous modules with a few learnable parameters are adopted which enable augmented paths to capture diverse information in low-dimensional subspaces. Since the parameters in the augmented paths can be reparametrized to the pre-trained model after fine-tuning, the proposed method incurs no additional inference cost. Extensive experiments and comparisons conducted on various visual benchmark tasks demonstrate the effectiveness of the proposed PPA method.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"703 ","pages":"Article 121948"},"PeriodicalIF":8.1,"publicationDate":"2025-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143349369","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

An improved grey wolf optimizer with flexible crossover and mutation for cluster task scheduling

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2025-02-06 DOI: 10.1016/j.ins.2025.121943

Hongbo Wang , Jinyu Zhang , Jingkun Fan , ChiYiDuo Zhang , Bo Deng , WenTao Zhao

With the rapid advancement of cloud computing, task scheduling algorithms inspired by natural phenomena have become a research focal point. The grey wolf optimizer (GWO), known for its strong convergence and ease of implementation, has attracted considerable attention. This study introduces an adaptive approach, GWO with the crossover and mutation variant (GWO_C/M), to integrate crossover and mutation strategies and thereby enhance the flexibility and applicability of the GWO. Rather than offering a fixed model, GWO_C/M employs different combinations of crossover and mutation strategies to enhance the balance between exploration and exploitation, solving issues including center bias. Extensive comparisons with 13 state-of-the-art (SOTA) models across six benchmark scenarios showed that GWO_C/M performed robustly, achieving an 87.2% success rate on 41 out of 47 test functions. Moreover, implementing GWO_C/M in CloudSim simulations markedly improved key performance metrics, including total execution time, task completion time, and load balancing. Further validation using the Alibaba Cluster Trace V2018 dataset confirmed that GWO_C/M improved resource utilization and reduced maximum task completion time, indicating the proposed approach's substantial benefits for task scheduling and overall system efficiency in cloud environments.

{"title":"An improved grey wolf optimizer with flexible crossover and mutation for cluster task scheduling","authors":"Hongbo Wang , Jinyu Zhang , Jingkun Fan , ChiYiDuo Zhang , Bo Deng , WenTao Zhao","doi":"10.1016/j.ins.2025.121943","DOIUrl":"10.1016/j.ins.2025.121943","url":null,"abstract":"<div><div>With the rapid advancement of cloud computing, task scheduling algorithms inspired by natural phenomena have become a research focal point. The grey wolf optimizer (GWO), known for its strong convergence and ease of implementation, has attracted considerable attention. This study introduces an adaptive approach, GWO with the crossover and mutation variant (GWO_C/M), to integrate crossover and mutation strategies and thereby enhance the flexibility and applicability of the GWO. Rather than offering a fixed model, GWO_C/M employs different combinations of crossover and mutation strategies to enhance the balance between exploration and exploitation, solving issues including center bias. Extensive comparisons with 13 state-of-the-art (SOTA) models across six benchmark scenarios showed that GWO_C/M performed robustly, achieving an 87.2% success rate on 41 out of 47 test functions. Moreover, implementing GWO_C/M in CloudSim simulations markedly improved key performance metrics, including total execution time, task completion time, and load balancing. Further validation using the Alibaba Cluster Trace V2018 dataset confirmed that GWO_C/M improved resource utilization and reduced maximum task completion time, indicating the proposed approach's substantial benefits for task scheduling and overall system efficiency in cloud environments.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"704 ","pages":"Article 121943"},"PeriodicalIF":8.1,"publicationDate":"2025-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143378819","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

GRVFL-MV: Graph random vector functional link based on multi-view learning

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2025-02-06 DOI: 10.1016/j.ins.2025.121947

M. Tanveer, R.K. Sharma , M. Sajid, A. Quadir

The classification performance of the random vector functional link (RVFL), a randomized neural network, has been widely acknowledged. However, due to its shallow learning nature, RVFL often fails to consider all the relevant information available in a dataset. Additionally, it overlooks the geometrical properties of the dataset. To address these limitations, a novel graph random vector functional link based on multi-view learning (GRVFL-MV) model is proposed. The proposed model is trained on multiple views, incorporating the concept of multiview learning (MVL), and also incorporates the geometrical properties of all the views using the graph embedding (GE) framework. The fusion of RVFL networks, MVL, and GE framework enables our proposed model to achieve the following: i) efficient learning: by leveraging the topology of RVFL, our proposed model can efficiently capture nonlinear relationships within the multi-view data, facilitating efficient and accurate predictions; ii) comprehensive representation: fusing information from diverse perspectives enhance the proposed model's ability to capture complex patterns and relationships within the data, thereby improving the model's overall generalization performance; and iii) structural awareness: by employing the GE framework, our proposed model leverages the original data distribution of the dataset by naturally exploiting both intrinsic and penalty subspace learning criteria. The evaluation of the proposed GRVFL-MV model on various datasets, including 29 UCI and KEEL datasets, 50 datasets from Corel5k, and 45 datasets from AwA, demonstrates its superior performance compared to baseline models. These results highlight the enhanced generalization capabilities of the proposed GRVFL-MV model across a diverse range of datasets. The source code of the proposed GRVFL-MV model is available at https://github.com/mtanveer1/GRVFL-MV.

{"title":"GRVFL-MV: Graph random vector functional link based on multi-view learning","authors":"M. Tanveer, R.K. Sharma , M. Sajid, A. Quadir","doi":"10.1016/j.ins.2025.121947","DOIUrl":"10.1016/j.ins.2025.121947","url":null,"abstract":"<div><div>The classification performance of the random vector functional link (RVFL), a randomized neural network, has been widely acknowledged. However, due to its shallow learning nature, RVFL often fails to consider all the relevant information available in a dataset. Additionally, it overlooks the geometrical properties of the dataset. To address these limitations, a novel graph random vector functional link based on multi-view learning (GRVFL-MV) model is proposed. The proposed model is trained on multiple views, incorporating the concept of multiview learning (MVL), and also incorporates the geometrical properties of all the views using the graph embedding (GE) framework. The fusion of RVFL networks, MVL, and GE framework enables our proposed model to achieve the following: i) <em>efficient learning</em>: by leveraging the topology of RVFL, our proposed model can efficiently capture nonlinear relationships within the multi-view data, facilitating efficient and accurate predictions; ii) <em>comprehensive representation</em>: fusing information from diverse perspectives enhance the proposed model's ability to capture complex patterns and relationships within the data, thereby improving the model's overall generalization performance; and iii) <em>structural awareness</em>: by employing the GE framework, our proposed model leverages the original data distribution of the dataset by naturally exploiting both intrinsic and penalty subspace learning criteria. The evaluation of the proposed GRVFL-MV model on various datasets, including 29 UCI and KEEL datasets, 50 datasets from Corel5k, and 45 datasets from AwA, demonstrates its superior performance compared to baseline models. These results highlight the enhanced generalization capabilities of the proposed GRVFL-MV model across a diverse range of datasets. The source code of the proposed GRVFL-MV model is available at <span><span>https://github.com/mtanveer1/GRVFL-MV</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"704 ","pages":"Article 121947"},"PeriodicalIF":8.1,"publicationDate":"2025-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143378821","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

Harnessing logic heterograph learning for financial operational risks: A perspective of cluster and thin-tailed distributions

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2025-02-05 DOI: 10.1016/j.ins.2025.121939

Guanyuan Yu, Boyu Han, Qing Li, Jiwen Huang

Financial operational risks, alongside credit and market risks, represent a primary concern for commercial banks. However, the inherent complexity of these risks, coupled with the challenges in defining and labeling operational activities, has constrained data-driven analysis in this domain. This study introduces a deep learning framework incorporating operational logic into risk identification through a heterograph embedding network (HEN). The HEN effectively condenses complex, high-dimensional operational logic heterographs into a more manageable low-dimensional space. Within this simplified space, a Density Estimation Network (DEN) is employed to pinpoint risks, drawing on the clustering properties and thin-tailed characteristics of financial data. These components are harmoniously combined through a unified objective function, designed to optimize both dimensionality reduction and classification decisions. Experimental evaluation on a real-world financial dataset reveals that the proposed framework surpasses several cutting-edge algorithms in terms of both practicality and effectiveness. More importantly, our approach is generalizable and can be extended to learn the logical connections between features across various domains.

{"title":"Harnessing logic heterograph learning for financial operational risks: A perspective of cluster and thin-tailed distributions","authors":"Guanyuan Yu, Boyu Han, Qing Li, Jiwen Huang","doi":"10.1016/j.ins.2025.121939","DOIUrl":"10.1016/j.ins.2025.121939","url":null,"abstract":"<div><div>Financial operational risks, alongside credit and market risks, represent a primary concern for commercial banks. However, the inherent complexity of these risks, coupled with the challenges in defining and labeling operational activities, has constrained data-driven analysis in this domain. This study introduces a deep learning framework incorporating operational logic into risk identification through a heterograph embedding network (HEN). The HEN effectively condenses complex, high-dimensional operational logic heterographs into a more manageable low-dimensional space. Within this simplified space, a Density Estimation Network (DEN) is employed to pinpoint risks, drawing on the clustering properties and thin-tailed characteristics of financial data. These components are harmoniously combined through a unified objective function, designed to optimize both dimensionality reduction and classification decisions. Experimental evaluation on a real-world financial dataset reveals that the proposed framework surpasses several cutting-edge algorithms in terms of both practicality and effectiveness. More importantly, our approach is generalizable and can be extended to learn the logical connections between features across various domains.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"704 ","pages":"Article 121939"},"PeriodicalIF":8.1,"publicationDate":"2025-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143377336","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

GMR-Rec: Graph mutual regularization learning for multi-domain recommendation

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2025-02-05 DOI: 10.1016/j.ins.2025.121946

Yifan Wang , Yangzi Yang , Shuai Li , Yutao Xie , Zhiping Xiao , Ming Zhang , Wei Ju

Multi-domain recommender systems are becoming increasingly significant, as they can alleviate the sparsity challenge and cold-start problem within a single domain by transferring knowledge from related domains in a collective manner. However, existing methods primarily concentrate on the process of sharing or mapping the features of the same users across different domains to facilitate knowledge transfer. Since the user-item interactions can be naturally formulated as bipartite graphs, transferring knowledge via message passing throughout domains would be a more straightforward approach. Moreover, the existing approaches generally pay more attention to modeling the common interests of users, leaving behind the under-explored domain-specific interests. In this paper, we introduce a novel framework, called GMR-Rec, for the multi-domain recommendation, which explicitly transfers knowledge across various domains. Specifically, both domain-shared and domain-specific graphs are constructed using historical user-item interactions, with the parallel graph neural network employed for each of them. Then, mutual regularization strategies are proposed to distinguish domain-specific user interests while preserving common user interests shared across domains. Experimental results on the four real-world datasets show that our model achieves an average improvement of 1.24%, 2.90%, 5.07% and 3.17% in HR@10, and 3.05%, 4.24%, 6.38% and 3.99% in NDCG@10 compared to the state-of-the-art baseline.

{"title":"GMR-Rec: Graph mutual regularization learning for multi-domain recommendation","authors":"Yifan Wang , Yangzi Yang , Shuai Li , Yutao Xie , Zhiping Xiao , Ming Zhang , Wei Ju","doi":"10.1016/j.ins.2025.121946","DOIUrl":"10.1016/j.ins.2025.121946","url":null,"abstract":"<div><div>Multi-domain recommender systems are becoming increasingly significant, as they can alleviate the sparsity challenge and cold-start problem within a single domain by transferring knowledge from related domains in a collective manner. However, existing methods primarily concentrate on the process of sharing or mapping the features of the same users across different domains to facilitate knowledge transfer. Since the user-item interactions can be naturally formulated as bipartite graphs, transferring knowledge via message passing throughout domains would be a more straightforward approach. Moreover, the existing approaches generally pay more attention to modeling the common interests of users, leaving behind the under-explored domain-specific interests. In this paper, we introduce a novel framework, called GMR-Rec, for the multi-domain recommendation, which explicitly transfers knowledge across various domains. Specifically, both domain-shared and domain-specific graphs are constructed using historical user-item interactions, with the parallel graph neural network employed for each of them. Then, mutual regularization strategies are proposed to distinguish domain-specific user interests while preserving common user interests shared across domains. Experimental results on the four real-world datasets show that our model achieves an average improvement of 1.24%, 2.90%, 5.07% and 3.17% in HR@10, and 3.05%, 4.24%, 6.38% and 3.99% in NDCG@10 compared to the state-of-the-art baseline.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"703 ","pages":"Article 121946"},"PeriodicalIF":8.1,"publicationDate":"2025-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143293053","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