Future Generation Computer Systems-The International Journal of Escience最新文献

{"title":"Data distribution aware clustering for parallel split learning in healthcare applications","authors":"Md. Tanvir Arafat ,&nbsp;Md. Abdur Razzaque ,&nbsp;Abdulhameed Alelaiwi ,&nbsp;Md. Zia Uddin ,&nbsp;Mohammad Mehedi Hassan","doi":"10.1016/j.future.2025.107911","DOIUrl":"10.1016/j.future.2025.107911","url":null,"abstract":"<div><div>Split learning, a promising approach in privacy-preserving machine learning, decentralizes model training by dividing it among client devices and a central server. However, split learning has exhibited a certain level of slowness in its vanilla approach, mainly due to the serial processing of devices. Recent research endeavors have addressed this challenge by introducing parallelism and thus accelerating the split learning process. However, the existing split learning methodologies often overlook the critical aspect of data distribution among client devices.</div><div>This paper introduces a Data Distribution Aware Clustering-based Parallel Split Learning (DCSL), a scheme purposefully crafted to address the complexities stemming from non-identically and non-independently distributed (non-IID) data among client devices engaged in the split learning paradigm. In healthcare applications, comprehending the intricacies of data distribution is imperative, particularly given the non-IID nature of medical datasets, to ensure accurate analysis and decision-making. The DCSL leverages a novel clustering technique to create clusters of medical client devices, considering the data distributions of their local datasets, and employs parallel model training within the device clusters. It enhances model convergence and reduces training latency by optimizing the cluster formation. Extensive experiments demonstrate that DCSL outperforms traditional split learning approaches, significantly improving accuracy and reducing training latency across various applications.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"174 ","pages":"Article 107911"},"PeriodicalIF":6.2,"publicationDate":"2025-06-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144305062","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Exploring RISC-V long vector capabilities: A case study in Earth Sciences","authors":"Fabio Banchelli,&nbsp;David Jurado,&nbsp;Marta Garcia-Gasulla,&nbsp;Filippo Mantovani","doi":"10.1016/j.future.2025.107932","DOIUrl":"10.1016/j.future.2025.107932","url":null,"abstract":"<div><div>This paper investigates the performance of Earth Sciences codes, specifically SeisSol and MiniFALL3D, on a RISC-V-based CPU featuring a long vector processing unit. The study focuses on optimizing these applications for improved computational efficiency while maintaining portability. For SeisSol, we explore batched GEMM implementations to enhance performance by leveraging instruction-level parallelism. MiniFALL3D’s optimization involves improving vectorization by modifying the source code, such as replacing functions with subroutines and flattening multidimensional arrays. The vectorization process is always left to the compiler to ensure code portability. The study is conducted using both a software emulator and a hardware prototype of the RISC-V vector architecture called EPAC. The performance of both applications is evaluated across different HPC platforms, including EPAC (based on RISC-V), MareNostrum 4 (powered by Sapphire Rapids CPUs), and the NEC SX-Aurora Tsubasa accelerator. We aim to provide insights into adapting Earth Sciences codes for modern high-performance computing systems while demonstrating the potential of RISC-V vector architectures. Ultimately, all modifications made to improve performance on the RISC-V long vector architecture are shown to be beneficial on other HPC architectures with different vector capabilities. This highlights the importance of maintaining code portability while relying on the compiler’s powerful auto-vectorization capabilities.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"174 ","pages":"Article 107932"},"PeriodicalIF":6.2,"publicationDate":"2025-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144262985","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Cooperative Task Offloading Strategy for Vehicular Edge Computing Based on Multi-Agent Deep Reinforcement Learning","authors":"Yuya Cui ,&nbsp;Degan Zhang ,&nbsp;Honghu Li ,&nbsp;Hao Qiang ,&nbsp;Haitao Zhao","doi":"10.1016/j.future.2025.107950","DOIUrl":"10.1016/j.future.2025.107950","url":null,"abstract":"<div><div>Vehicular Edge Computing (VEC) and Vehicle- to-Vehicle (V2V) offloading can significantly reduce in-vehicle task latency. This paper investigates a cooperative task offloading strategy in VEC, where latency-sensitive and computation -intensive tasks can be offloaded to Road Side Units (RSUs) using 5 G connectivity. Additionally, these tasks can be shared among nearby vehicles through V2V links. Joint VEC and V2V cooperative offloading can not only minimizes task execution delays but also prevents network congestion. When vehicles are in motion, dynamic migration of computation tasks is necessary to maintain service continuity. We propose a two-phase distributed task offloading and migration strategy for multiple vehicles. In the first phase, vehicles select the optimal service vehicle based on inter-vehicle link quality and offloading willingness. In the second phase, to minimize system cost, we introduce a multi-agent reinforcement learning (MARL) based distributed task offloading and migration strategy. This strategy allows vehicles to choose the optimal edge node in a dynamic environment without fully offloading information. Moreover, we implement a counterfactual multi- agent (COMA) reinforcement learning approach to address the inefficiency caused by the credit allocation problem in multi-agent systems. Extensive evaluations demonstrate that the algorithm proposed in this paper perform better in terms of average system latency and overall task completion rate. Compared with related scheme, the proposed method can reduce latency by up to 54 % and improve task completion rate by up to 15 % in different scenarios.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"174 ","pages":"Article 107950"},"PeriodicalIF":6.2,"publicationDate":"2025-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144262987","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multi-omic and quantum machine learning integration for lung subtypes classification","authors":"Mandeep Kaur Saggi ,&nbsp;Amandeep Singh Bhatia ,&nbsp;Isaiah K. Mensah ,&nbsp;Humaira Gowher ,&nbsp;Sabre Kais","doi":"10.1016/j.future.2025.107905","DOIUrl":"10.1016/j.future.2025.107905","url":null,"abstract":"<div><div>The integration of multi-omics data presents a promising frontier in cancer diagnosis and biomarker discovery, especially for complex diseases like lung cancer. However, challenges such as high dimensionality, low sample sizes, and inherent data noise hinder traditional machine-learning approaches. Quantum Machine Learning (QML) is a cutting-edge field that bridges quantum computing and machine learning to address computational challenges more effectively. This study explores the application of QML to address these limitations, offering a novel framework-Multi-Omic QML Lung Subtype Classification (MQML-LungSC)-for classifying lung cancer subtypes: lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC). Leveraging Quantum Neural Networks with multi-dimensional feature encoding, our model efficiently integrates genomic, epigenomic, and transcriptomic data from TCGA. The model not only achieves high classification accuracy (training: 0.95; testing: 0.90) using 256 encoded features, but also demonstrates enhanced efficiency by outperforming classical machine learning methods and other quantum models with a significantly reduced architectural complexity. Notably, QNN-64,delivers performance comparable to CNN-64 while maintaining a more compact and resource-efficient design. By identifying key differentiating features, this approach advances early diagnostic capabilities and supports personalized treatment strategies. This study provides strong empirical support for the future potential of unconventional computing approaches in advancing biomedical research and applications.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"174 ","pages":"Article 107905"},"PeriodicalIF":6.2,"publicationDate":"2025-06-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144240645","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"TAP: Distributed team assignment in heterogeneous multi-agent systems","authors":"Deniz Ozsoyeller","doi":"10.1016/j.future.2025.107925","DOIUrl":"10.1016/j.future.2025.107925","url":null,"abstract":"<div><div>In this article, we introduce and study the problem of autonomous balanced team assignment in a heterogeneous multi-robot (i.e. multi-agent) system. The system includes <span><math><mi>n</mi></math></span> robots that are initially located in a large open area. We consider a scenario where there are two types of robots, namely, worker and service, with limited communication ranges and specialized capabilities. The robots should be divided into teams so that each team can be assigned to a different location. The objective is to minimize the maximum distance traveled among the worker robots while ensuring that each constructed team is of equal size and has exactly one service robot to assist the worker robots in the team. Depending on the robot’s initial configuration, the robot can be either single or a part of a connected communication network of robots. In the former case, the robot does not know the location of any other robot, whereas in the latter case, the robot only knows the locations of the robots in its network but not the ones outside it. For this problem, we propose two algorithms, TA_m and TA_nm, that combine the distributed coordination and online motion planning methods. For assignment, TA_m uses a mutual decision making approach, whereas TA_nm uses a nonmutual decision making approach. We evaluate the performances of our strategies through extensive simulations varying the key parameters of interest including communication range, environment size, number of worker robots, and number of service robots. The results show that TA_m outperforms TA_nm in sparse configurations, but the performances of the algorithms approach to each other as the configuration becomes dense.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"174 ","pages":"Article 107925"},"PeriodicalIF":6.2,"publicationDate":"2025-06-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144230571","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Practical volume-hiding range searchable symmetric encryption using trusted execution","authors":"Xu Yang,&nbsp;Ke Li,&nbsp;Saiyu Qi,&nbsp;Hongguang Zhao","doi":"10.1016/j.future.2025.107930","DOIUrl":"10.1016/j.future.2025.107930","url":null,"abstract":"<div><div>Searchable symmetric encryption (SSE) enables clients to securely outsource private data to cloud servers while preserving search functionality. Addressing the increasing demand for range queries, which involve matching documents to fields with a range of keywords, remains a challenge. Existing volume-hiding range query schemes fail to address volume pattern leakage during the document fetch phase and lack support for dynamic operations such as addition and deletion of documents. To overcome these limitations, we propose EDVHRQ, an efficient, dynamic, and volume-hiding range query scheme enabled by Intel SGX. Our scheme introduces novel strategies: (1) packetization, which conceals response sizes while minimizing server storage and communication costs, and (2) the optimal best range cover (OBRC) method, which transforms query ranges into a minimal set of trapdoors to accelerate range queries. Unlike existing solutions, EDVHRQ achieves a single roundtrip query process between the client and the cloud server. We formally analyze the security guarantees of EDVHRQ, demonstrating its robustness against volume pattern leakage. Experimental evaluations highlight its superior performance, achieving up to 18.8<span><math><mo>×</mo></math></span> and 56.5<span><math><mo>×</mo></math></span> <em>faster</em> query execution compared to HybrIDX and SEAL, respectively, while significantly reducing server storage costs.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"174 ","pages":"Article 107930"},"PeriodicalIF":6.2,"publicationDate":"2025-06-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144230424","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A novel nonparametric Bayesian model for time series clustering: Application to electricity load profile characterization","authors":"Lincheng Han,&nbsp;Shun Cheng,&nbsp;Weiru Yuan,&nbsp;Xiuyu Zhang,&nbsp;Jianguo Wang","doi":"10.1016/j.future.2025.107929","DOIUrl":"10.1016/j.future.2025.107929","url":null,"abstract":"<div><div>Accurate characterization of electricity load profiles is essential for energy management and power grid optimization. This paper introduces a novel time series clustering model, the Dirichlet Process Mixture of Sparse Heteroskedastic Multi-output Gaussian Processes (DPM-SHMGP), for enhanced load profile characterization and consumption pattern identification. The model employs the Dirichlet process for automatic cluster number determination and integrates the Gaussian process to capture time-varying dependencies in the data while providing uncertainty quantification for clustering results. This methodology overcomes several limitations of conventional clustering algorithms when processing temporal data, such as handling temporal misalignment, processing incomplete data, and distinguishing between sequences that share identical probability distributions but exhibit different temporal patterns. The approach incorporates a Linear Model of Coregionalization for joint modeling of data means and standard deviations while implementing a Subset of Data approximation method to mitigate the computational burden of Gaussian processes. Experimental results on synthetic and real-world datasets demonstrate the model’s superior clustering performance compared to established clustering algorithms while providing reliable cluster number estimates and accurate load profile characterization. Additional experiments validate the model’s optimal cluster number determination capability and demonstrate its performance stability across data conditions, including subset sizes and missing values.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"174 ","pages":"Article 107929"},"PeriodicalIF":6.2,"publicationDate":"2025-06-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144312603","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Hierarchical and distributed data storage for computing continuum","authors":"Elias Del-Pozo-Puñal,&nbsp;Felix Garcia-Carballeira,&nbsp;Diego Camarmas-Alonso,&nbsp;Alejandro Calderon-Mateos","doi":"10.1016/j.future.2025.107931","DOIUrl":"10.1016/j.future.2025.107931","url":null,"abstract":"<div><div>The Internet of Things (IoT) has transformed how everyone interacts with the environment. Over the past few years, this field has experienced exponential growth, which has led to difficulties in efficiently managing the data generated by these devices and has posed new challenges for cloud infrastructures. As the number of devices involved increases, latency and bandwidth issues become increasingly critical in these systems. To address these issues, architectures such as fog and edge computing emerged that proposed bringing information processing and storage closer to the data generators. This reduced the distance the data had to travel, thereby improving latency and bandwidth and reducing potential bottlenecks.</div><div>These architectures have evolved into a new concept known as the computing continuum. This approach, based on dynamic collaboration between the cloud, fog, and edge, creates a continuous infrastructure of computational resources that optimizes data processing at each network level as needed. The computing continuum presents important challenges that need to be addressed at multiple levels: the application/algorithmic level (programming paradigms), middleware level (deployment, execution, scheduling, monitoring, data storage, transfer, processing, and analysis), and resource management level.</div><div>The work introduced in this article addresses the challenges related to the efficient storage and transfer of data between the different levels across the computing continuum. We propose a distributed and parallel file system for this kind of infrastructure that can be used transparently at all levels. This file system can be deployed at different levels (fog, edge, and cloud) hierarchically, allowing the execution of applications at each level and efficient data transfer between these levels and the cloud. It also facilitates the development of IoT applications capable of efficiently transferring data by using typical file system calls.</div><div>The work presents and validates this data storage system at different levels: two IoT devices (Raspberry Pi 1 &amp; 4), an emulated environment, a controlled simulation, and a performance analysis with Amazon Cloud Services.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"174 ","pages":"Article 107931"},"PeriodicalIF":6.2,"publicationDate":"2025-06-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144230577","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"BCE-PPDS: Blockchain-based cloud–edge collaborative privacy-preserving data sharing scheme for IoT","authors":"Guijuan Wang ,&nbsp;Qi Liu ,&nbsp;Zhongyuan Yu ,&nbsp;Hongliang Zhang ,&nbsp;Anming Dong","doi":"10.1016/j.future.2025.107922","DOIUrl":"10.1016/j.future.2025.107922","url":null,"abstract":"<div><div>Internet of Things (IoT) devices generate large amounts of data every day that can be combined with intelligent platforms for predictive analytics and scientific research. However, concerns about privacy and security hinder the willingness of individuals to share data. Blockchain emerged as a promising infrastructure for facilitating secure data sharing due to its decentralized, immutability, and auditable benefits. In this paper, we propose a blockchain-based cloud–edge collaborative privacy protection data sharing scheme (BCE-PPDS), which is decentralized and enables data requesters (DRs) to search data resources using smart contracts to efficiently obtain target data. To protect the identity privacy of data owners (DOs), we propose a novel certificateless linkable ring signature algorithm with efficient performance. This algorithm is not only suitable for deployment on resource-limited IoT devices, so that DOs can realize anonymous identity authentication, but also can aggregate the generated ring signatures for batch verification, so as to improve the efficiency of signature verification. In addition, we designed a key distribution algorithm using the Asmuth–Bloom secret sharing scheme to ensure the security of the key. Under the random oracle model, BCE-PPDS is provably secure. The experimental results verify that BCE-PPDS is efficient and practical.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"174 ","pages":"Article 107922"},"PeriodicalIF":6.2,"publicationDate":"2025-05-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144195238","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"State of practice: Evaluating GPU performance of state vector and tensor network methods","authors":"Marzio Vallero,&nbsp;Paolo Rech,&nbsp;Flavio Vella","doi":"10.1016/j.future.2025.107927","DOIUrl":"10.1016/j.future.2025.107927","url":null,"abstract":"<div><div>The frontier of quantum computing (QC) simulation on classical hardware is quickly reaching the hard scalability limits for computational feasibility. Nonetheless, there is still a need to simulate large quantum systems classically, as the Noisy Intermediate Scale Quantum (NISQ) devices are yet to be considered fault tolerant and performant enough in terms of operations per second. Each of the two main exact simulation techniques, state vector and tensor network simulators, boasts specific limitations.</div><div>This article investigates the limits of current state-of-the-art simulation techniques on a test bench made of eight widely used quantum subroutines, each in different configurations, with a special emphasis on performance. We perform both single process and distributed scaleability experiments on a supercomputer. We correlate the performance measures from such experiments with the metrics that characterise the benchmark circuits, identifying the main reasons behind the observed performance trends. Specifically, we perform distributed sliced tensor contractions, and we analyse the impact of pathfinding quality on contraction time, correlating both results with topological circuit characteristics. From our observations, given the structure of a quantum circuit and the number of qubits, we highlight how to select the best simulation strategy, demonstrating how preventive circuit analysis can guide and improve simulation performance by more than an order of magnitude.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"174 ","pages":"Article 107927"},"PeriodicalIF":6.2,"publicationDate":"2025-05-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144223669","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}