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

{"title":"DP-SWAP: Fast Swapping Strategy Based on Dynamic Programming","authors":"Weiduo Chen ,&nbsp;Xiaoshe Dong ,&nbsp;Qiang Wang","doi":"10.1016/j.future.2025.108071","DOIUrl":"10.1016/j.future.2025.108071","url":null,"abstract":"<div><div>Neural Architecture Search (NAS) has emerged as an effective approach for automating neural network design. However, NAS imposes significant GPU memory pressure due to the need to evaluate numerous candidate models during training. While tensor swapping helps reduce memory usage, existing tensor selection methods rely on extensive iterative searches, which require repeatedly traversing model computation graphs to evaluate the impact of swapping schemes–leading to high time complexity and poor scalability in dynamic NAS scenarios.</div><div>To address this issue, we propose DP-SWAP, a novel tensor swapping strategy based on dynamic programming. By leveraging the optimal substructure property of the tensor selection problem, DP-SWAP computes effective swapping schemes with only <span><math><mrow><mi>O</mi><mo>(</mo><mi>n</mi><mo>)</mo></mrow></math></span> time complexity, allows for fast and adaptive decision-making during NAS model exploration.</div><div>Experimental results show that DP-SWAP achieves training performance comparable to state-of-the-art heuristic methods, while reducing swapping decision time by over 3 orders of magnitude, thus effectively alleviating GPU memory bottlenecks in NAS.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"175 ","pages":"Article 108071"},"PeriodicalIF":6.2,"publicationDate":"2025-08-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144889177","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":"Correction is all you need: Towards high-order complementary cloud API recommendation correction with abductive reasoning","authors":"Zhen Chen ,&nbsp;Haonan Liao ,&nbsp;Jingkun Yang ,&nbsp;Mengyao Wu ,&nbsp;Dianlong You","doi":"10.1016/j.future.2025.108072","DOIUrl":"10.1016/j.future.2025.108072","url":null,"abstract":"<div><div>In the cloud era, cloud Application Programming Interfaces (APIs) are the best carriers for service delivery, data exchange, and capability replication. The continuously growing number of cloud APIs in dynamic open networks provides developers with more choices but also overwhelms them with a vast array of options. Existing researches primarily use invocation preferences, search keywords, and quality of service modeling to generate a single function cloud API recommendation list. However, these methods face two problems: (1) In service-oriented software development, developers’ needs for high-order complementary cloud APIs are often overlooked. (2) Current cloud API recommendation methods generate recommendation results in a one-shot manner without further correcting them to enhance performance. To tackle these problems, we propose high-order complementary cloud API recommendation correction with abductive reasoning, named HCCR-CAR. HCCR-CAR comprises of two stages: HCCR and CAR. In HCCR, the complementary scores of candidate cloud APIs is determined by taking into account both the explicit and fine-grained complementary relationships between the cloud API query set and the candidate APIs. Subsequently, the candidate cloud APIs are ranked based on these complementary scores in order to generate high-order complementary recommendation results. In CAR, a reasonable abductive task is designed and an abductive model is utilized to infer the most probable complementary reason for the recommendation results produced by HCCR. By minimizing the abductive loss signal between inferred reason and real reason through back-propagation, the recommendation results are corrected. Experiments are conducted on two real-world cloud API datasets which demonstrate that compared to traditional heuristic recommendation methods and deep learning recommendation methods, HCCR-CAR exhibits superior performance in high-order complementary cloud API recommendation. Furthermore, hyperparameter sensitivity experiments and case analysis validate the effectiveness and practicality of this method. HCCR-CAR is more likely to yield satisfactory results for developers while also ensuring the effectiveness of high-order complementary cloud API recommendation in practical service-oriented software development, thereby effectively enhancing revenue of cloud API providers.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"175 ","pages":"Article 108072"},"PeriodicalIF":6.2,"publicationDate":"2025-08-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144864307","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":"PRepChain: A versatile privacy-preserving reputation system for dynamic supply chain environments","authors":"Jan Pennekamp ,&nbsp;Lennart Bader ,&nbsp;Emildeon Thevaraj ,&nbsp;Stefanie Berninger ,&nbsp;Martin Perau ,&nbsp;Tobias Schröer ,&nbsp;Wolfgang Boos ,&nbsp;Salil S. Kanhere ,&nbsp;Klaus Wehrle","doi":"10.1016/j.future.2025.108024","DOIUrl":"10.1016/j.future.2025.108024","url":null,"abstract":"<div><div>Despite their significant added value in the context of consumer-oriented e-commerce, reputation systems have seen limited adoption in other business settings and models these days. Yet, reliable reputation scores are essential in such settings for easing the establishment of new business relationships—an aspect that is particularly crucial in dynamic supply chain environments, where business partners change frequently. Existing approaches, however, usually target other application domains and fall short in addressing the specific challenges of dynamic supply chains—especially with respect to reliability (incl. availability) and privacy preservation (incl. confidentiality). To close this research gap and to support novel directions in this important research area, we propose PRepChain, our highly-configurable approach that leverages fully homomorphic encryption and distributed competences to provide businesses with a versatile reputation-enriched ecosystem. PRepChain is specifically designed to operate in dynamic environments by also offering a trade-off between data availability and confidentiality guarantees. We make contributions in four primary directions: (i) It offers performant privacy preservation even in large-scale settings, (ii) ensures availability of computed reputation scores, (iii) seamlessly integrates with existing supply chain information systems, and (iv) in addition to subjective reputation scores, it also supports reliably-calculated, i.e., objective, ones, thereby strengthening the reliability of third-party-sourced information. Our evaluation of PRepChain documents its performance—based on a real-world use case—, security, and privacy preservation, hence, its applicability. We conclude that it is indeed destined for practical deployments in modern supply networks.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"175 ","pages":"Article 108024"},"PeriodicalIF":6.2,"publicationDate":"2025-08-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144858054","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":"Distributed scalable multi-agent reinforcement learning with intrinsic-episodic dual exploration","authors":"Shuhan Qi,&nbsp;Shuhao Zhang,&nbsp;Qiang Wang,&nbsp;Jiajia Zhang,&nbsp;Xuan Wang","doi":"10.1016/j.future.2025.108040","DOIUrl":"10.1016/j.future.2025.108040","url":null,"abstract":"<div><div>Cooperative multi-agent reinforcement learning still faces challenges in multi-agent exploration and data-efficiency. In this paper, we propose a practical framework named Distributed Scalable Multi-Agent Reinforcement Learning with Intrinsic-Episodic Dual Exploration (SIEMA) to tackle these challenges. Under the widely-applied assumption of centralized training with decentralized execution and value decomposition assumption, SIEMA encourages multi-agent exploration and addresses the issue of low sample utilization through Intrinsic-Episodic Dual Exploration. The Cooperative Exploration Intrinsic Reward (CEIR) component incentivizes exploration from various aspects, incorporating novelty, optimal distance, and cooperative exploration. Episodic Exploration Replay (EER) explores at the episode level, ensuring optimal utilization of all samples in the replay buffer. Furthermore, we introduce the distributed scalable multi-agent training framework to accelerate the learning process and address the issue of low sample generation in MARL by deploying multiple workers and actors in a distributed manner. We illustrate the advantages of SIEMA by ablation experiments, and demonstrate its remarkable superiority over state-of-the-art MARL algorithms on challenging tasks in the StarCraft II micromanagement benchmark.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"175 ","pages":"Article 108040"},"PeriodicalIF":6.2,"publicationDate":"2025-08-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144840882","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":"Mosaic: Composite projection pruning for resource-efficient LLMs","authors":"Bailey J. Eccles ,&nbsp;Leon Wong ,&nbsp;Blesson Varghese","doi":"10.1016/j.future.2025.108056","DOIUrl":"10.1016/j.future.2025.108056","url":null,"abstract":"<div><div>Extensive compute and memory requirements limit the deployment of large language models (LLMs) on any hardware. Compression methods, such as pruning, can reduce model size, which in turn reduces resource requirements. State-of-the-art pruning is based on coarse-grained methods. They are time-consuming and inherently remove critical model parameters, adversely impacting the quality of the pruned model. This paper introduces projection pruning, a novel fine-grained method for pruning LLMs. In addition, LLM projection pruning is enhanced by a new approach we refer to as composite projection pruning — the synergistic combination of unstructured pruning that retains accuracy and structured pruning that reduces model size. We develop <span><span>Mosaic</span></span>, a novel system to create and deploy pruned LLMs using composite projection pruning. <span><span>Mosaic</span></span> is evaluated using a range of performance and quality metrics on multiple hardware platforms, LLMs, and datasets. <span><span>Mosaic</span></span> is 7.19<span><math><mo>×</mo></math></span> faster in producing models than existing approaches. <span><span>Mosaic</span></span> models achieve up to 84.2% lower perplexity and 31.4% higher accuracy than models obtained from coarse-grained pruning. Up to 67% faster inference and 68% lower GPU memory use is noted for <span><span>Mosaic</span></span> models.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"175 ","pages":"Article 108056"},"PeriodicalIF":6.2,"publicationDate":"2025-08-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144864152","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":"Supervised contrastive learning-based stress detection for wearable sensor-based healthcare applications","authors":"Kyungtaek Oh ,&nbsp;Hyunseo Park ,&nbsp;Gyeong Ho Lee ,&nbsp;Jun Kyun Choi","doi":"10.1016/j.future.2025.108058","DOIUrl":"10.1016/j.future.2025.108058","url":null,"abstract":"<div><div>Advancements in Internet of Things (IoT) technology have enabled continuous physiological monitoring through wearable devices, significantly improving personalized healthcare services. Automated stress detection based on physiological signals is a critical function in digital health systems, enabling timely interventions and reducing severe health risks. Despite recent progress in deep learning, effectively capturing both general human stress patterns and individual-specific variations remains a core challenge. To address this, we propose StressCon, a novel deep learning framework that fully integrates contrastive learning with metadata fusion to learn robust stress patterns while adapting to individual characteristics. Furthermore, our method jointly optimizes two contrastive loss functions to learn subject-invariant features, and then leverages user metadata for personalized stress detection. We comprehensively evaluate the proposed method on two publicly available datasets using Leave-One-Subject-Out (LOSO) validation. Results show that StressCon enhances classification accuracy by up to 3.49% and F1-score by up to 3.48%, while maintaining consistent state-of-the-art performance across diverse populations. These findings confirm the superior generalization capabilities and practical applicability of the proposed approach, demonstrating an effective balance between population-level robustness and individual personalization for IoT-based stress monitoring systems.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"175 ","pages":"Article 108058"},"PeriodicalIF":6.2,"publicationDate":"2025-08-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144830713","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":"Assuming the best: Towards a reliable protocol for resource usage prediction for high-performance computing based on machine learning","authors":"Alexandre H.L. Porto ,&nbsp;Micaella Coelho ,&nbsp;Hiago M.G.A. Rocha ,&nbsp;Carla Osthoff ,&nbsp;Kary Ocaña ,&nbsp;Douglas O. Cardoso","doi":"10.1016/j.future.2025.108070","DOIUrl":"10.1016/j.future.2025.108070","url":null,"abstract":"<div><div>In High-Performance Computing (HPC) systems, multiple processes simultaneously consume resources such as CPU time, memory, and electrical power, among others. Accurately predicting the resource consumption of a process based on its execution parameters enables more efficient resource allocation, ultimately improving the overall performance of the HPC system. While many studies have explored this topic, fewer explicitly examine the underlying assumptions of their approaches. This work contributes to filling that gap by proposing, experimenting with, and discussing a protocol to approach this problem, covering from the collection of processes footprint data to the experimental evaluation of Machine Learning models based on such data. The reported results of the assessment of this protocol in a case study of the RAxML bioinformatics application on a real supercomputer highlight not only its effectiveness (<span><math><msup><mi>R</mi><mn>2</mn></msup></math></span> values greater than 0.9 were achieved in most tests) but also the reasonableness of the assumptions considered.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"175 ","pages":"Article 108070"},"PeriodicalIF":6.2,"publicationDate":"2025-08-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144864151","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":"On-device derivation of IoT usage control policies: Automating U-XACML policy generation from natural language with LLMs in smart homes environments","authors":"Loay Alajramy ,&nbsp;Marco Simoni ,&nbsp;Marco Rasori ,&nbsp;Andrea Saracino ,&nbsp;Paolo Mori","doi":"10.1016/j.future.2025.108067","DOIUrl":"10.1016/j.future.2025.108067","url":null,"abstract":"<div><div>In this paper, we present a framework that integrates AI-based derivation of Access and Usage Control policies for IoT devices, using Large Language Models (LLMs) to automate the generation of policies from unstructured natural language commands. The framework employs a hybrid approach, combining LLMs with dedicated libraries to ensure efficient on-device execution. Our approach is based on a two-step process: first, a fine-tuned LLM converts user commands into structured JSON policy representations; then, a transformation module translates the JSON policies into fully compliant U-XACML policies. To ensure generality across different domains, we introduce a taxonomy-driven dataset creation, which enables policy creation for different environments such as smart homes, smart offices, and healthcare settings. Our evaluation demonstrates that the system achieves 93 % accuracy in policy generation and 91 % accuracy when handling ambiguous or noisy inputs. It also reaches 98 % agreement with expert-defined policies in real-world scenarios. Finally, on-device performance evaluations confirm the feasibility of running the model in practical settings, demonstrating reliable inference under constrained hardware conditions.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"175 ","pages":"Article 108067"},"PeriodicalIF":6.2,"publicationDate":"2025-08-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144885611","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 function-as-a-service middleware for decentralized collaborative edge computing","authors":"Catarina Gonçalves ,&nbsp;José Simão ,&nbsp;Luís Veiga","doi":"10.1016/j.future.2025.108069","DOIUrl":"10.1016/j.future.2025.108069","url":null,"abstract":"<div><div>Function-as-a-Service (FaaS) emerges as a sophisticated cloud computing paradigm critically suited to processing the exponentially increasing data volumes generated by Internet of Things (IoT) infrastructures. Deploying computational models proximal to data generation sources addresses critical latency and bandwidth constraints inherent in edge-distributed applications. Edge computing environments present complex architectural challenges characterized by large-scale decentralized infrastructures and resource-constrained devices, which substantially impede contemporary Function-as-a-Service implementation strategies. This research introduces FaaS@Edge, a novel framework that leverages volunteered edge node resources discovered through the InterPlanetary File System (IPFS) network and deployed via Apache OpenWhisk. The proposed system addresses computational resource distribution challenges by enabling FaaS runtime deployments across heterogeneous edge infrastructure. Comprehensive experimental evaluation shows that FaaS@Edge introduces marginal latency during function submission while maintaining performance comparable to local OpenWhisk implementations. Empirical results demonstrate request success rates that approximate 98 % for function submission and invocation processes. These findings shows FaaS@Edge’s potential as an efficient computational model for edge computing environments, characterized by low-latency performance and optimized resource allocation.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"175 ","pages":"Article 108069"},"PeriodicalIF":6.2,"publicationDate":"2025-08-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144864153","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":"Optimizing CPU-GPU resource scheduling with deep reinforcement learning","authors":"Kaijie Chen ,&nbsp;Kai Huang ,&nbsp;Jian Mao ,&nbsp;Jiawei Hu ,&nbsp;Jinliang Lin ,&nbsp;Zhengxian You","doi":"10.1016/j.future.2025.108065","DOIUrl":"10.1016/j.future.2025.108065","url":null,"abstract":"<div><div>The rapid development of the Internet of Things (IoT) and Edge Computing has created a greater need for real-time and accurate information updates in latency-sensitive applications. We address heterogeneous devices, including CPU-only, GPU-only, and hybrid CPU-GPU devices, by constructing a Markov Decision Process (MDP) model that effectively captures the dynamic characteristics of these devices and Edge Server (ES). To fully leverage the heterogeneity of wireless devices(WDs), we propose a Deep Reinforcement Learning (DRL) algorithm based on a Multi-Head Attention, where each device type is assigned independent attention weights to capture its impact on task scheduling decisions efficiently. To address the slow convergence issue in traditional Reinforcement Learning (RL) algorithms under completely unknown systems, we propose heterogeneous computing-aware Post-Decision States (PDS) learning. This mechanism incorporates partial prior knowledge of the dynamics of the system in edge environments to accelerate the exploration and learning process. Experimental results demonstrate that the proposed method significantly optimizes both the Age of Information (AoI) and the energy consumption performance in heterogeneous edge environments, outperforming existing approaches.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"175 ","pages":"Article 108065"},"PeriodicalIF":6.2,"publicationDate":"2025-08-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144880010","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}