Pub Date : 2024-12-31DOI: 10.1016/j.future.2024.107683
Bin Wang , Zhao Tian , Jie Ma , Wenju Zhang , Wei She , Wei Liu
The traditional synchronous federated learning framework ensures global model consistency and accuracy. However, it is limited by the computational power differences between devices and the influence of non-IID data, which leads to inefficient training and insufficient model generalization performance. In this paper, we propose a decentralized asynchronous federated learning framework. The framework uses smart contracts deployed on the blockchain to manage edge devices for enhanced flexibility. At first, the framework performs model aggregation and validation through the use of consensus groups. It eliminates the potential single point of failure associated with centralized parameter servers. In addition, we propose a Federated Learning with Dynamically Growing Cache (FedDgc) method in a non-IID environment. The method reduces redundant gradient information exchange during initial feature extraction while maintaining the learning capability of the global model. Finally, the experimental results show that our framework has better test performance and guarantees the convergence speed of the model during training.
{"title":"A decentralized asynchronous federated learning framework for edge devices","authors":"Bin Wang , Zhao Tian , Jie Ma , Wenju Zhang , Wei She , Wei Liu","doi":"10.1016/j.future.2024.107683","DOIUrl":"10.1016/j.future.2024.107683","url":null,"abstract":"<div><div>The traditional synchronous federated learning framework ensures global model consistency and accuracy. However, it is limited by the computational power differences between devices and the influence of non-IID data, which leads to inefficient training and insufficient model generalization performance. In this paper, we propose a decentralized asynchronous federated learning framework. The framework uses smart contracts deployed on the blockchain to manage edge devices for enhanced flexibility. At first, the framework performs model aggregation and validation through the use of consensus groups. It eliminates the potential single point of failure associated with centralized parameter servers. In addition, we propose a Federated Learning with Dynamically Growing Cache (FedDgc) method in a non-IID environment. The method reduces redundant gradient information exchange during initial feature extraction while maintaining the learning capability of the global model. Finally, the experimental results show that our framework has better test performance and guarantees the convergence speed of the model during training.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"166 ","pages":"Article 107683"},"PeriodicalIF":6.2,"publicationDate":"2024-12-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143166081","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}
Pub Date : 2024-12-31DOI: 10.1016/j.future.2024.107695
Lance Drane, Marshall McDonnell, Randall Petras, Cody Stiner, Arthur J. Ruckman, Gavin M. Wiggins, Gregory Cage, Robert Smith, Seth Hitefield, Jesse McGaha, Andrew Ayres, Mike Brim, Richard Archibald, Addi Malviya-Thakur
In the rapidly evolving field of frontend development, Single-Page Applications (SPAs) stand out for their ability to create dynamic and interactive web applications, particularly valuable in scientific software for their real-time data integration and complex workflow management. However, the process of creating a single-page web application development environment that accurately reflects the production environment isn’t always straightforward. Most SPA build systems assume configuration at build time, while DevSecOps engineers prefer runtime configuration. This paper proposes a unique, framework-agnostic methodology designed to bridge this divide, facilitating the seamless integration of SPAs within the DevSecOps framework without necessitating expertise in both domains. Leveraging environmental variables, Docker, and a strategic approach to Content Security Policy (CSP), we provide a comprehensive guide for developing, deploying, and securing SPAs in a manner that is both efficient and secure. Applying this method to the INTERSECT and Smart Spectral Matching platforms, we demonstrate its effectiveness in enhancing both the development process and the user experience in scientific applications, thereby addressing the complex challenges faced by research software engineers in the current landscape.
{"title":"Integrating scientific single-page applications with DevSecOps","authors":"Lance Drane, Marshall McDonnell, Randall Petras, Cody Stiner, Arthur J. Ruckman, Gavin M. Wiggins, Gregory Cage, Robert Smith, Seth Hitefield, Jesse McGaha, Andrew Ayres, Mike Brim, Richard Archibald, Addi Malviya-Thakur","doi":"10.1016/j.future.2024.107695","DOIUrl":"10.1016/j.future.2024.107695","url":null,"abstract":"<div><div>In the rapidly evolving field of frontend development, Single-Page Applications (SPAs) stand out for their ability to create dynamic and interactive web applications, particularly valuable in scientific software for their real-time data integration and complex workflow management. However, the process of creating a single-page web application development environment that accurately reflects the production environment isn’t always straightforward. Most SPA build systems assume configuration at build time, while DevSecOps engineers prefer runtime configuration. This paper proposes a unique, framework-agnostic methodology designed to bridge this divide, facilitating the seamless integration of SPAs within the DevSecOps framework without necessitating expertise in both domains. Leveraging environmental variables, Docker, and a strategic approach to Content Security Policy (CSP), we provide a comprehensive guide for developing, deploying, and securing SPAs in a manner that is both efficient and secure. Applying this method to the INTERSECT and Smart Spectral Matching platforms, we demonstrate its effectiveness in enhancing both the development process and the user experience in scientific applications, thereby addressing the complex challenges faced by research software engineers in the current landscape.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"166 ","pages":"Article 107695"},"PeriodicalIF":6.2,"publicationDate":"2024-12-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143166796","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}
The Internet of Things (IoT) and Edge-Cloud Computing have been trending technologies over the past few years. In this work, we introduce the Enhanced Optimized-Greedy Nominator Heuristic (EO-GNH), a framework designed to optimize machine learning (ML) and artificial intelligence (AI) application placement in edge environments, aiming to improve Quality of Service (QoS). Developed specifically for sectors such as smart agriculture, industry, and healthcare, EO-GNH integrates asynchronous MapReduce and parallel meta-heuristics to effectively manage AI applications, focusing on execution performance, resource utilization, and infrastructure resilience. The framework carefully addresses the distribution challenges of AI applications, especially Service Function Chains (SFCs), in edge-cloud infrastructures. It contains Data Flow Management, which covers aspects of data storage and data privacy, and also considers factors like regional adaptations, mobile access, and AI model refinement. EO-GNH ensures high availability for forecasting, prediction, and training AI models, operating efficiently within a geo-distributed infrastructure. The proposed strategies within EO-GNH emphasize concurrent multi-node execution, enhancing AI application placement by improving execution time, dependability, and cost-effectiveness. The efficiency of EO-GNH is demonstrated through its impact on QoS in real-time resource management across three application domains, highlighting its adaptability and potential in diverse cross-domain IoT-based environments.
{"title":"Enhancing performance of machine learning tasks on edge-cloud infrastructures: A cross-domain Internet of Things based framework","authors":"Osama Almurshed , Ashish Kaushal , Souham Meshoul , Asmail Muftah , Osama Almoghamis , Ioan Petri , Nitin Auluck , Omer Rana","doi":"10.1016/j.future.2024.107696","DOIUrl":"10.1016/j.future.2024.107696","url":null,"abstract":"<div><div>The Internet of Things (IoT) and Edge-Cloud Computing have been trending technologies over the past few years. In this work, we introduce the Enhanced Optimized-Greedy Nominator Heuristic (EO-GNH), a framework designed to optimize machine learning (ML) and artificial intelligence (AI) application placement in edge environments, aiming to improve Quality of Service (QoS). Developed specifically for sectors such as smart agriculture, industry, and healthcare, EO-GNH integrates asynchronous MapReduce and parallel meta-heuristics to effectively manage AI applications, focusing on execution performance, resource utilization, and infrastructure resilience. The framework carefully addresses the distribution challenges of AI applications, especially Service Function Chains (SFCs), in edge-cloud infrastructures. It contains Data Flow Management, which covers aspects of data storage and data privacy, and also considers factors like regional adaptations, mobile access, and AI model refinement. EO-GNH ensures high availability for forecasting, prediction, and training AI models, operating efficiently within a geo-distributed infrastructure. The proposed strategies within EO-GNH emphasize concurrent multi-node execution, enhancing AI application placement by improving execution time, dependability, and cost-effectiveness. The efficiency of EO-GNH is demonstrated through its impact on QoS in real-time resource management across three application domains, highlighting its adaptability and potential in diverse cross-domain IoT-based environments.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"166 ","pages":"Article 107696"},"PeriodicalIF":6.2,"publicationDate":"2024-12-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143167490","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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}
Pub Date : 2024-12-30DOI: 10.1016/j.future.2024.107694
Jianguo Liang , Qianqian Li , Hao Han , You Fu
Molecular dynamics (MD) simulation, a crucial technique for investigating atomic structure and dynamic properties, has become a primary method for studying the thermodynamic properties of dielectric materials, such as silicon, and their low-dimensional nanostructures. Diamond-structured semiconductors exhibit unique crystallographic properties. Achieving optimal simulation performance on supercomputing platforms necessitates specialized parallel design and optimization, considering both atom spatial characteristics and platform architecture. To tackle storage challenges in large-scale simulations of diamond-structured crystals, we designed a hierarchical storage-based atom data organization and a neighbor list construction algorithm exploiting positional offsets. Furthermore, a novel “point-line-plane” communication model was implemented. This model leverages the distribution of atom neighbors and a fixed neighbor list, enhancing communication efficiency via data packing to enable scalable simulations. A numerical simulation software, Diamond-MD, was developed for simulating diamond-structured crystals, enabling simulations at the 100 million-atom scale. Benchmark results indicate that Diamond-MD achieves a 44% reduction in memory usage and a 48% improvement in computational performance compared to LAMMPS. Moreover, Diamond-MD demonstrates excellent scalability.
{"title":"Parallel software design of large-scale diamond-structured crystals molecular dynamics simulation","authors":"Jianguo Liang , Qianqian Li , Hao Han , You Fu","doi":"10.1016/j.future.2024.107694","DOIUrl":"10.1016/j.future.2024.107694","url":null,"abstract":"<div><div>Molecular dynamics (MD) simulation, a crucial technique for investigating atomic structure and dynamic properties, has become a primary method for studying the thermodynamic properties of dielectric materials, such as silicon, and their low-dimensional nanostructures. Diamond-structured semiconductors exhibit unique crystallographic properties. Achieving optimal simulation performance on supercomputing platforms necessitates specialized parallel design and optimization, considering both atom spatial characteristics and platform architecture. To tackle storage challenges in large-scale simulations of diamond-structured crystals, we designed a hierarchical storage-based atom data organization and a neighbor list construction algorithm exploiting positional offsets. Furthermore, a novel “point-line-plane” communication model was implemented. This model leverages the distribution of atom neighbors and a fixed neighbor list, enhancing communication efficiency via data packing to enable scalable simulations. A numerical simulation software, Diamond-MD, was developed for simulating diamond-structured crystals, enabling simulations at the 100 million-atom scale. Benchmark results indicate that Diamond-MD achieves a 44% reduction in memory usage and a 48% improvement in computational performance compared to LAMMPS. Moreover, Diamond-MD demonstrates excellent scalability.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"166 ","pages":"Article 107694"},"PeriodicalIF":6.2,"publicationDate":"2024-12-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143167345","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}
Pub Date : 2024-12-30DOI: 10.1016/j.future.2024.107693
Yiran Li , Yankun Cao , Jia Mi , Xiaoxiao Cui , Xifeng Hu , Yuezhong Zhang , Zhi Liu , Lizhen Cui , Shuo Li
The successful recognition of the standard echocardiographic ten-views remains elusive, primarily due to the complexity of cardiac anatomy, confusion caused by low-quality data, and subtle variations among closely related multi-views. To cope with the limitations of existing algorithms, which include a lack of objectivity, accuracy, and robustness, we propose a Hybrid Cooperative Metric Network (HCMN). We enhance the objectivity, accuracy and robustness of the quality assessment by integrating knowledge of cycle-consistency with metric consistency, which helps mitigate inaccurate fitting in hybrid distributions. Therefore, it provides a clear feature similarity distribution to prevent feature confusion. The experiments demonstrate that the HCMN model significantly outperforms the state-of-the-art in quality assessment, achieving an impressive accuracy of 96.74%. We believe this novel framework will establish a reliable benchmark for recognizing standard echocardiographic multi-views and provide a new interpretable perspective on standardized the automatic cardiac disease diagnosis. By adapting and applying advanced assessment methodologies, we can enhance the clarity and interpretability of medical imaging, thereby aiding in the precise identification of lesions and improving decision-making accuracy in drug discovery.
{"title":"Cooperative metric learning-based hybrid transformer for automatic recognition of standard echocardiographic multi-views","authors":"Yiran Li , Yankun Cao , Jia Mi , Xiaoxiao Cui , Xifeng Hu , Yuezhong Zhang , Zhi Liu , Lizhen Cui , Shuo Li","doi":"10.1016/j.future.2024.107693","DOIUrl":"10.1016/j.future.2024.107693","url":null,"abstract":"<div><div>The successful recognition of the standard echocardiographic ten-views remains elusive, primarily due to the complexity of cardiac anatomy, confusion caused by low-quality data, and subtle variations among closely related multi-views. To cope with the limitations of existing algorithms, which include a lack of objectivity, accuracy, and robustness, we propose a Hybrid Cooperative Metric Network (HCMN). We enhance the objectivity, accuracy and robustness of the quality assessment by integrating knowledge of cycle-consistency with metric consistency, which helps mitigate inaccurate fitting in hybrid distributions. Therefore, it provides a clear feature similarity distribution to prevent feature confusion. The experiments demonstrate that the HCMN model significantly outperforms the state-of-the-art in quality assessment, achieving an impressive accuracy of 96.74%. We believe this novel framework will establish a reliable benchmark for recognizing standard echocardiographic multi-views and provide a new interpretable perspective on standardized the automatic cardiac disease diagnosis. By adapting and applying advanced assessment methodologies, we can enhance the clarity and interpretability of medical imaging, thereby aiding in the precise identification of lesions and improving decision-making accuracy in drug discovery.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"166 ","pages":"Article 107693"},"PeriodicalIF":6.2,"publicationDate":"2024-12-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142929251","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}
Pub Date : 2024-12-28DOI: 10.1016/j.future.2024.107686
R. Zhang, W. Ni, N. Fu, L. Hou, D. Zhang, Y. Zhang
Trajectory data has a wide range of applications in various domains but also raises serious privacy concerns. To address these concerns, the integration of deep learning with differential privacy for trajectory publication has gained widespread attention. However, existing solutions are mostly based on temporal neural networks and the Generative Adversarial Networks (GAN) framework, which intrinsically faces the ”forgetting” problem, leading to their failure to capture and simulate the multi-time-scale behavioral patterns of traffic participants, thereby reducing the utility of published trajectories. Moreover, the computational cost and the noise introduced by the widely used differentially private gradient perturbation method are proportional to the model size, which compromises model quality. To address these problems, we propose a Differentially Private Trajectory Publishing method via Locally-aware Transformer-based GAN (DP-LTGAN), achieving high-utility trajectory publishing while providing differential privacy protection. Specifically, our method features a Locally-aware Transformer, whose attention mechanism is refined by incorporating local state encoding and a multi-scale temporal encoding mechanism. This enhancement significantly improves the modeling of both long- and short-term trajectory patterns. Furthermore, a differentially private gradient perturbation method named Common Term Perturbation (CTP) has been developed, which effectively reduces the amount of noise and the computational cost by utilizing a designed local noise addition pattern and an adaptive noise addition mechanism. Extensive experiments on several real trajectory datasets show that our method enhances the utility and efficiency of synthetic trajectories by 57.7% and 46.88%, respectively, significantly outperforming current state-of-the-art approaches.
{"title":"DP-LTGAN: Differentially private trajectory publishing via Locally-aware Transformer-based GAN","authors":"R. Zhang, W. Ni, N. Fu, L. Hou, D. Zhang, Y. Zhang","doi":"10.1016/j.future.2024.107686","DOIUrl":"10.1016/j.future.2024.107686","url":null,"abstract":"<div><div>Trajectory data has a wide range of applications in various domains but also raises serious privacy concerns. To address these concerns, the integration of deep learning with differential privacy for trajectory publication has gained widespread attention. However, existing solutions are mostly based on temporal neural networks and the Generative Adversarial Networks (GAN) framework, which intrinsically faces the ”forgetting” problem, leading to their failure to capture and simulate the multi-time-scale behavioral patterns of traffic participants, thereby reducing the utility of published trajectories. Moreover, the computational cost and the noise introduced by the widely used differentially private gradient perturbation method are proportional to the model size, which compromises model quality. To address these problems, we propose a Differentially Private Trajectory Publishing method via Locally-aware Transformer-based GAN (DP-LTGAN), achieving high-utility trajectory publishing while providing differential privacy protection. Specifically, our method features a Locally-aware Transformer, whose attention mechanism is refined by incorporating local state encoding and a multi-scale temporal encoding mechanism. This enhancement significantly improves the modeling of both long- and short-term trajectory patterns. Furthermore, a differentially private gradient perturbation method named Common Term Perturbation (CTP) has been developed, which effectively reduces the amount of noise and the computational cost by utilizing a designed local noise addition pattern and an adaptive noise addition mechanism. Extensive experiments on several real trajectory datasets show that our method enhances the utility and efficiency of synthetic trajectories by 57.7% and 46.88%, respectively, significantly outperforming current state-of-the-art approaches.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"166 ","pages":"Article 107686"},"PeriodicalIF":6.2,"publicationDate":"2024-12-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143166204","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}
Pub Date : 2024-12-28DOI: 10.1016/j.future.2024.107689
José A. González-Nóvoa , Laura Busto , Silvia Campanioni , Carlos Martínez , José Fariña , Juan J. Rodríguez-Andina , Pablo Juan-Salvadores , Víctor Jiménez , Andrés Íñiguez , César Veiga
Hypertension remains a leading cause of premature mortality globally, emphasizing the critical need for early detection and management. Unfortunately, less than half of hypertensive adults receive proper diagnosis and treatment. To address this gap, continuous blood pressure (ABP) monitoring has emerged as a valuable tool for detecting cardiovascular complications before they escalate. ABP monitoring can be achieved by cuffless ABP estimation techniques embedded on wearables. In this paper, we present an innovative personalized medicine approach for cuffless arterial blood pressure estimation, characterized by its patient-specific focus and computational requirements reduction. An XGBoost patient specific ABP estimator model is optimized for each patient through Bayesian techniques, using their photoplethysmogram (PPG) features. The proposed method achieves a mean absolute error (MAE) of 7.27 mmHg for systolic and 3.33 mmHg for diastolic blood pressure. Additionally, recursive feature elimination techniques are used to streamline the model, making it suitable for resource-limited environments such as wearables platforms. This combination of approaches offers a promising outlook for the application of personalized medicine in blood pressure monitoring, thereby enhancing hypertension management and reducing associated health risks.
{"title":"Advancing cuffless arterial blood pressure estimation: A patient-specific optimized approach reducing computational requirements","authors":"José A. González-Nóvoa , Laura Busto , Silvia Campanioni , Carlos Martínez , José Fariña , Juan J. Rodríguez-Andina , Pablo Juan-Salvadores , Víctor Jiménez , Andrés Íñiguez , César Veiga","doi":"10.1016/j.future.2024.107689","DOIUrl":"10.1016/j.future.2024.107689","url":null,"abstract":"<div><div>Hypertension remains a leading cause of premature mortality globally, emphasizing the critical need for early detection and management. Unfortunately, less than half of hypertensive adults receive proper diagnosis and treatment. To address this gap, continuous blood pressure (ABP) monitoring has emerged as a valuable tool for detecting cardiovascular complications before they escalate. ABP monitoring can be achieved by cuffless ABP estimation techniques embedded on wearables. In this paper, we present an innovative personalized medicine approach for cuffless arterial blood pressure estimation, characterized by its patient-specific focus and computational requirements reduction. An XGBoost patient specific ABP estimator model is optimized for each patient through Bayesian techniques, using their photoplethysmogram (PPG) features. The proposed method achieves a mean absolute error (MAE) of 7.27 mmHg for systolic and 3.33 mmHg for diastolic blood pressure. Additionally, recursive feature elimination techniques are used to streamline the model, making it suitable for resource-limited environments such as wearables platforms. This combination of approaches offers a promising outlook for the application of personalized medicine in blood pressure monitoring, thereby enhancing hypertension management and reducing associated health risks.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"166 ","pages":"Article 107689"},"PeriodicalIF":6.2,"publicationDate":"2024-12-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143167346","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}
Pub Date : 2024-12-24DOI: 10.1016/j.future.2024.107676
Xubin Wang, Wenju Li, Chang Lu
Expert in pattern recognition, deep learning is widely used for defect detection. Data augmentation is in principle capable of improving the accuracy and robustness of such data-driven models by supplementing patterns. However, the requirements for realistic images, precise annotations, and diverse defect patterns often cannot be addressed simultaneously in current augmentation methods. In this work, a Mask Guided Cross Data Augmentation method dubbed MGCDA using diffusion model is proposed to boost defect detection. Firstly, a generation pipeline in latent diffusion space utilizing autoencoder is formulated to improve the fidelity and resource effort. Based on this, we propose to adopt conditional mechanism to enable samples being synthesized under the guidance of specific masks. To further enhance the information gain, a cross-learning strategy is proposed to empower MGCDA learning and generalizing diverse defect patterns from different categories, making detection more robust. Finally, two strategies are proposed to tackle the demand for data augmentation in different situations. Experiments on eight common industrial datasets show that MGCDA has high applicability to different scenarios and detection models, it can generate high-fidelity samples aligned to guidance and effectively improve the performance of baselines at both image- and pixel-level.
{"title":"A mask guided cross data augmentation method for industrial defect detection","authors":"Xubin Wang, Wenju Li, Chang Lu","doi":"10.1016/j.future.2024.107676","DOIUrl":"10.1016/j.future.2024.107676","url":null,"abstract":"<div><div>Expert in pattern recognition, deep learning is widely used for defect detection. Data augmentation is in principle capable of improving the accuracy and robustness of such data-driven models by supplementing patterns. However, the requirements for realistic images, precise annotations, and diverse defect patterns often cannot be addressed simultaneously in current augmentation methods. In this work, a <strong>M</strong>ask <strong>G</strong>uided <strong>C</strong>ross <strong>D</strong>ata <strong>A</strong>ugmentation method dubbed MGCDA using diffusion model is proposed to boost defect detection. Firstly, a generation pipeline in latent diffusion space utilizing autoencoder is formulated to improve the fidelity and resource effort. Based on this, we propose to adopt conditional mechanism to enable samples being synthesized under the guidance of specific masks. To further enhance the information gain, a cross-learning strategy is proposed to empower MGCDA learning and generalizing diverse defect patterns from different categories, making detection more robust. Finally, two strategies are proposed to tackle the demand for data augmentation in different situations. Experiments on eight common industrial datasets show that MGCDA has high applicability to different scenarios and detection models, it can generate high-fidelity samples aligned to guidance and effectively improve the performance of baselines at both image- and pixel-level.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"166 ","pages":"Article 107676"},"PeriodicalIF":6.2,"publicationDate":"2024-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142929269","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}
Pub Date : 2024-12-24DOI: 10.1016/j.future.2024.107677
Jingtong Huang , Xu Ma , Yuan Ma , Kehao Chen , Xiaoyu Zhang
Graph neural networks (GNNs) are effective for graph-based node classification tasks, such as data mining and recommendation systems. Combining federated learning(FL) with GNN enables multiple participants to collaboratively train powerful models without sharing private data. However, subgraph-level FL faces challenges, including missing cross-client edges and non-IID data distributions. Additionally, ensuring security in non-fully trusted environments is a critical concern. To address these issues, we propose RMFGL (Robust Meta Federated Graph Learning), a framework for subgraph-level node classification. RMFGL integrates cross-client information through pre-feature aggregation and leverages model-agnostic meta-learning (MAML) to optimize meta-parameters with minimal federated updates. For robustness, we employ a GCN architecture with dual-channel attention aggregation, while Multi-key Fully Homomorphic Encryption (MKFHE) ensures privacy during training. Experimental results on Cora, CiteSeer, PubMed and Coauthor-CS datasets show that RMFGL achieves up to a 2x accuracy improvement with minimal fine-tuning compared to baseline methods and outperforms state-of-the-art techniques. Notably, RMFGL significantly enhances robustness against malicious clients, with up to 100x improvement in stability, while maintaining strong performance with non-IID data.
{"title":"Dual-channel meta-federated graph learning with robust aggregation and privacy enhancement","authors":"Jingtong Huang , Xu Ma , Yuan Ma , Kehao Chen , Xiaoyu Zhang","doi":"10.1016/j.future.2024.107677","DOIUrl":"10.1016/j.future.2024.107677","url":null,"abstract":"<div><div>Graph neural networks (GNNs) are effective for graph-based node classification tasks, such as data mining and recommendation systems. Combining federated learning(FL) with GNN enables multiple participants to collaboratively train powerful models without sharing private data. However, subgraph-level FL faces challenges, including missing cross-client edges and non-IID data distributions. Additionally, ensuring security in non-fully trusted environments is a critical concern. To address these issues, we propose RMFGL (<strong>R</strong>obust <strong>M</strong>eta <strong>F</strong>ederated <strong>G</strong>raph <strong>L</strong>earning), a framework for subgraph-level node classification. RMFGL integrates cross-client information through pre-feature aggregation and leverages model-agnostic meta-learning (MAML) to optimize meta-parameters with minimal federated updates. For robustness, we employ a GCN architecture with dual-channel attention aggregation, while Multi-key Fully Homomorphic Encryption (MKFHE) ensures privacy during training. Experimental results on Cora, CiteSeer, PubMed and Coauthor-CS datasets show that RMFGL achieves up to a 2x accuracy improvement with minimal fine-tuning compared to baseline methods and outperforms state-of-the-art techniques. Notably, RMFGL significantly enhances robustness against malicious clients, with up to 100x improvement in stability, while maintaining strong performance with non-IID data.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"166 ","pages":"Article 107677"},"PeriodicalIF":6.2,"publicationDate":"2024-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142889237","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}
Pub Date : 2024-12-21DOI: 10.1016/j.future.2024.107678
Hongzhi Xu , Binlian Zhang , Chen Pan , Keqin Li
With the rapid development of information technology, efficient energy utilization has become a major challenge in modern computing system design. This paper focuses on the energy-constrained parallel application scheduling problem in heterogeneous systems and proposes three algorithms to minimize the makespan of applications. The first one is the minimum makespan algorithm under energy constraints. In this algorithm, we construct an optimal cost table with energy constraints, which can be applied to determine the priority of tasks and the processors allocated in the application. The second one is the energy reclaiming algorithm, which is used to reclaim some energy from non-critical tasks while ensuring that the makespan of the application remains unchanged. The third one is the energy reallocation algorithm, which tends to allocate reclaimed energy to critical tasks to increase their execution frequency, thereby reducing the makespan of the entire application. Experiments were conducted on different parallel applications in various scenarios, and the results showed that the proposed algorithm can achieve smaller makespan compared to existing algorithms in most cases.
{"title":"Scheduling energy-constrained parallel applications in heterogeneous systems","authors":"Hongzhi Xu , Binlian Zhang , Chen Pan , Keqin Li","doi":"10.1016/j.future.2024.107678","DOIUrl":"10.1016/j.future.2024.107678","url":null,"abstract":"<div><div>With the rapid development of information technology, efficient energy utilization has become a major challenge in modern computing system design. This paper focuses on the energy-constrained parallel application scheduling problem in heterogeneous systems and proposes three algorithms to minimize the makespan of applications. The first one is the minimum makespan algorithm under energy constraints. In this algorithm, we construct an optimal cost table with energy constraints, which can be applied to determine the priority of tasks and the processors allocated in the application. The second one is the energy reclaiming algorithm, which is used to reclaim some energy from non-critical tasks while ensuring that the makespan of the application remains unchanged. The third one is the energy reallocation algorithm, which tends to allocate reclaimed energy to critical tasks to increase their execution frequency, thereby reducing the makespan of the entire application. Experiments were conducted on different parallel applications in various scenarios, and the results showed that the proposed algorithm can achieve smaller makespan compared to existing algorithms in most cases.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"166 ","pages":"Article 107678"},"PeriodicalIF":6.2,"publicationDate":"2024-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142889238","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}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号