检索结果-内蒙古大学图书馆

continuous learning deraining network based on residual FFT convolution and contextual transformer module

IET IMAGE PROCESSING 2023年第3期17卷 747-760页

作者： Zhang, Zhijia Wu, Sinan Peng, Xinming Wang, Wanting Li, Rui Shenyang Univ Technol Sch Artificial Intelligence Shenyang 110178 Liaoning Peoples R China Shenyang Univ Technol Sch Informat Sci & Engn Shenyang Liaoning Peoples R China

Single image deraining work has attracted a lot of attention. So an effective single image deraining network model, RFCTNet, is designed. Specifically, the residual Fast Fourier Transform convolution module is used as a baseline for the deraining network. The advantage is its ability to capture both long and short-term information interactions and to integrate high and low-frequency background information for delivery. And a novel Transformer module, the Contextual Transformer module, is introduced. Contextual knowledge of the rain streaks neighbourhood is learned to improve the recovery of textures and information in the image background based on focusing on the rain pattern features. Because of the overlap between rain streaks, removing them in one stage is difficult, so a multi-stage progressive rain removal method is used. Considering the catastrophic forgetting problem of deep learning, a continuous learning method, PIGWM, is adopted, which is c and enables the network model to memorize of the previous data set. Also, the rain removal method RFCTNet has been proven through extensive experiments to outperform other advanced methods' evaluation metrics on both synthetic and real datasets. The source code can be found on

关键词： contextual transformer continuous learning image deraining residual FFT convolution module

来源：评论

学校读者我要写书评

暂无评论

continuous learning Method of Radar HRRP Based on CVAE-GAN

引用

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 2023年 61卷 1页

作者： Li, Xungen Ouyang, Wenqing Pan, Mian Lv, Shuaishuai Ma, Qi Hangzhou Dianzi Univ Sch Elect & Informat Hangzhou 310018 Peoples R China

To improve the catastrophic forgetting that existed in the high-resolution range profile (HRRP) target recognition model of offline training, this article proposed a continuous learning method of radar HRRP based on conditional variational auto-encoding and generative adversarial network (CVAE-GAN). First, this method generates data through CVAE to simulate the real training data and replays it in a series of subsequent tasks. By using generators, the proposed method does not need to save the original HRRP data and can ensure the privacy of the training dataset. Second, the proposed method takes the categorical data label as a generating condition, solves the unbalanced categories of the generated data samples, and improves the accuracy of the radar HRRP target recognition. Finally, by combining the attention mechanism and GAN network of the transformer, the generative capacity of the CVAE generator is enhanced effectively. In this article, the continuous learning method is verified by a validation framework with three task settings. The experimental results show that the proposed method has a better continuous learning ability and better engineering practicality in various tasks compared with the previous regularization methods which can guarantee the privacy of the training dataset.

关键词： Catastrophic forgetting conditional variational auto-encoding and generative adversarial network (CVAE-GAN) continuous learning high-resolution range profile (HRRP) radar target recognition

来源：评论

学校读者我要写书评

暂无评论

continuous learning in SMES: the mediating role of resilient dynamic capabilities across the firm life cycle

引用

learning ORGANIZATION 2025年第4期32卷 672-694页

作者： Quansah, Emmanuel Solansky, Stephanie Wang, Yuan Moghaddam, Kaveh Univ Houston Victoria Dept Management Operat & Mkt Victoria TX 77901 USA Case Western Reserve Univ Weatherhead Sch Management Cleveland OH 44106 USA Texas State Univ San Marcos Dept Management San Marcos TX USA

PurposeSmall and medium-sized enterprises (SMEs) are critical in the global business environment, yet they face significant challenges of survival. This study aims to investigate the role of continuous learning and resilience dynamic capabilities (RDC) on SME ***/methodology/approachThe authors conducted a quantitative survey of leaders from 290 SMEs to examine the relationship between continuous learning and performance. Structural equation modeling (SEM) is used to test the dependent relationships in this study. In addition, the Monte Carlo Method is used to test the mediating effect of *** results suggest that there is a direct effect of continuous learning on firm performance, as well as an indirect effect through RDC. In other words, the effect of continuous learning on firm performance was partially mediated through ***/valueThe authors also explored whether firm life cycle stage matters regarding the role of continuous learning and RDC in SMEs. The mediating role of RDC becomes more pronounced (fully mediated) as SMEs transition from the growth stage to maturity stage of their life cycle.

关键词： SMEs continuous learning Resilient dynamic capabilities Firm performance

来源：评论

学校读者我要写书评

暂无评论

MUFTI: Multi-Domain Distillation-Based Heterogeneous Federated continuous learning

引用

IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY 2025年 20卷 2721-2733页

作者： Gai, Keke Wang, Zijun Yu, Jing Zhu, Liehuang Beijing Inst Technol Sch Cyberspace Sci & Technol Beijing 100081 Peoples R China Minzu Univ China Sch Informat Engn Beijing 100081 Peoples R China

Federated learning (FL) is an alternative approach that facilitates training machine learning models on distributed users' data while preserving privacy. However, clients have different local model structures and most local data are non-independent and identically distributed, so that FL encounters heterogeneity and catastrophic forgetting issues when clients continuously accumulate new knowledge. In this work, we propose a scheme called MUFTI (Multi-Domain Distillation-based Heterogeneous Federated continuous learning). On one hand, we have extended domain adaptation to FL via extracting features to obtain feature representations on unlabeled public datasets for collaborative training, narrowing the distance between feature outputs of different models under the same sample. On the other hand, we propose a combining knowledge distillation method to solve the catastrophic forgetting issue. Within a single task, dual-domain distillation is used to avoid data forgetting between different domains;for cross task learning in task flow, the logits output of the previous model is used as the teacher to avoid forgetting old tasks. The experiment results showed that MUFTI had a better performance in accuracy and robustness comparing to state-of-the-art methods. The evaluation also demonstrated that MUFTI could perform well in handling task increment issues, reducing catastrophic forgetting, and achieving trade-offs between multiple objectives.

关键词： Heterogeneous federated learning continuous learning knowledge distillation Heterogeneous federated learning continuous learning knowledge distillation

来源：评论

学校读者我要写书评

暂无评论

Online Management for Edge-Cloud Collaborative continuous learning: A Two-Timescale Approach

引用

IEEE TRANSACTIONS ON MOBILE COMPUTING 2024年第12期23卷 14561-14574页

作者： Lin, Shaohui Zhang, Xiaoxi Li, Yupeng Joe-Wong, Carlee Duan, Jingpu Yu, Dongxiao Wu, Yu Chen, Xu Sun Yat Sen Univ Sch Comp Sci & Engn Guangzhou 510006 Peoples R China Hong Kong Baptist Univ Dept Interact Media Kowloon Tong Hong Kong Peoples R China Carnegie Mellon Univ Dept Elect & Comp Engn Pittsburgh PA 15213 USA Pengcheng Lab Dept Commun Shenzhen 518066 Peoples R China Shandong Univ Inst Intelligent Comp Sch Comp Sci & Technol Qingdao 266237 Peoples R China Dongguan Univ Technol Sch Cyberspace Secur Dongguan 523808 Peoples R China

Deep learning (DL) powered real-time applications usually need continuous training using data streams generated over time and across different geographical locations. Enabling data offloading among computation nodes through model training is promising to mitigate the problem that devices generating large datasets may have low computation capability. However, offloading can compromise model convergence and incur communication costs, which must be balanced with the long-term cost spent on computation and model synchronization. Therefore, this paper proposes EdgeC3, a novel framework that can optimize the frequency of model aggregation and dynamic offloading for continuously generated data streams, navigating the trade-off between long-term accuracy and cost. We first provide a new error bound to capture the impacts of data dynamics that are varying over time and heterogeneous across devices, as well as quantifying varied data heterogeneity between local models and the global one. Based on the bound, we design a two-timescale online optimization framework. We periodically learn the synchronization frequency to adapt with uncertain future offloading and network changes. In the finer timescale, we manage online offloading by extending Lyapunov optimization techniques to handle an unconventional setting, where our long-term global constraint can have abruptly changed aggregation frequencies that are decided in the longer timescale. Finally, we theoretically prove the convergence of EdgeC3 by integrating the coupled effects of our two-timescale decisions, and we demonstrate its advantage through extensive experiments performing distributed DL training for different domains.

关键词： Data models Training Computational modeling Accuracy Costs Distributed databases Federated learning Collaborative federated learning continuous learning edge-cloud collaboration two-timescale

来源：评论

学校读者我要写书评

暂无评论

continuous learning and cooperative prediction for traffic dynamics by Adaptive Multi-Agent Systems

Continuous learning and cooperative prediction for traffic d...

引用

4th IEEE International Conference on Autonomic Computing and Self-Organizing Systems (ACSOS)

作者： Ha Nhi Ngo Kaddoum, Elsy Gleizes, Marie-Pierre Bonnet, Jonathan Goursolle, Anais IRIT Comp Sci Res Inst Toulouse Toulouse France Continental Digital Serv France Toulouse France

ISBN: (纸本)9798350337440

Prediction of traffic dynamics plays a significant role in many Intelligent Transportation Systems (ITS). Nonetheless, accurate and real-time traffic prediction is always a difficult task. The classical models are challenged by the complex spatiotemporal relationships of the road network that raises unsolved questions relating the reliability and the feasibility of prediction models. Nowadays, the development of localization and communication technologies in transportation has led to massive data collected by on-board sensors known as floating car data (FCD). These data sets open up a new direction for traffic prediction using big data analysis methods. In this paper, we propose to address the traffic dynamics prediction problem using a self-adaptive multi-agent system that aims at continuously processing vehicle trajectory data to detect and learn different traffic dynamics and thus predict traffic evolution. The proposed system includes two processes: local learning, which distributes learning tasks at the agent level, and prediction process, which enables accurate traffic prediction using cooperative interactions among agents. The conducted experiments underline the performance of our system compared to the well-known models in the traffic prediction domain.

关键词： Traffic prediction dynamic clustering Multi-Agent System continuous learning self-adaptive mechanisms

来源：评论

学校读者我要写书评

暂无评论

Multivariate sensor data analysis for fault detection toward feedback loop-based continuous learning in FFF 3D printer

引用

INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING TECHNOLOGY 2024年第11-12期132卷 5975-5988页

作者： Baek, Sujeong Hanbat Natl Univ Dept Ind & Management Engn Room 211Bldg N4125 Dongseodae Ro Daejeon 34158 South Korea

Real-time fault detection via sensor signal analyses is widely performed in electromechanical systems to prevent sudden stops and the resulting losses, which requires a sufficient amount of fault data. Because it is difficult to obtain sufficient fault data with actual physical systems, fault detection is usually conducted using unsupervised detection models with datasets collected under health operations. Supervised models with few fault datasets of limited types have been developed, often with sophisticatedly augmented fault datasets. However, actual faults occur cumulatively with machine use, and the information contained in the actual data is very valuable. Therefore, this study proposes a feedback loop-based fault detection method that can be adjusted depending on the fault datasets during operation. Unsupervised fault detection is conducted using a long short-term memory autoencoder-based deep learning model and statistical process control chart with a derived anomaly score before observing any fault state. After obtaining even a small amount of fault datasets, fully connected layers are added for supervised classification, such as transfer learning techniques, to determine whether the current state is healthy or faulty. The performance of the proposed continuous learning method for fault detection is verified using datasets from a fused filament fabrication three-dimensional printer under several scenarios: no fault, one type of fault, and two types of faults. The detection performance improves as informative feedback is provided by the newly added layers. It was observed that even small fault datasets have the potential to improve the performance of trained unsupervised fault detection model. Since it is possible to detect fault occurrence from the initial state of the product life cycle, and the performance would be automatically updated as the printer is used, it is expected that it can be installed in a commercial three-dimensional printer and used direct

关键词： FFF 3D printer Fault detection Feedback loop Multivariate sensor data continuous learning

来源：评论

学校读者我要写书评

暂无评论

Human Activity Recognition System Based on continuous learning with Human Skeleton Information

引用

SENSORS AND MATERIALS 2024年第11期36卷 4713-4730页

作者： Dou, Wenbang Azhar, Aulia Saputra Chin, Weihong Kubota, Naoyuki Tokyo Metropolitan Univ Grad Sch Syst Design 6-6 Asahigaoka Hino Tokyo 1910065 Japan

In recent years, as the demographic profile of society continues to shift towards an aging population, there has been a concomitant shortage of caregivers, leading to an increase in the demand for elderly care. The accurate assessment of the health status of the elderly and the provision of appropriate care necessitate the timely recognition and analysis of human activities. To address this challenge, we propose a continuous human activity recognition system that generates a 3D human skeleton model, utilizes joint angles to perform daily life activity recognition, and infer similarities in movements across various body parts. The proposed system generates a 3D human skeleton model using depth information obtained from multiple range- based depth cameras and extracts human joint angles on the basis of this model. Moreover, it utilizes time-series joint angle data to continuously recognize actions and estimate the similarity of movements across various body parts. To validate the efficacy of the proposed system, comprehensive verification experiments were conducted using real-world data.

关键词： human skeleton model human activity recognition continuous learning

来源：评论

学校读者我要写书评

暂无评论

A self-organized consensus-reaching model considering personalized individual semantic continuous learning for LSGDM: application for medical consumables assessment

引用

INTERNATIONAL JOURNAL OF MACHINE learning AND CYBERNETICS 2025年 1-24页

作者： Yang, Chai Hu, Xiaoxuan Yu, Decai Wang, Yanjun Hefei Univ Technol Sch Management Hefei 230009 Peoples R China Univ Sci & Technol China Affiliated Hosp USTC 1 Div Life Sci & Med Hefei 230031 Peoples R China Anhui Jianzhu Univ Sch Math & Phys Hefei 230601 Peoples R China

Hospital-based Health Technology Assessment (HB-HTA) requires active stakeholder participation and significant financial investment. Selecting the appropriate healthcare assessment scheme and reaching consensus are essential. Given the diversity in educational backgrounds, knowledge, and experiences, HB-HTA participants often rely on linguistic information to express their individual preferences. However, since word meanings can vary among individuals, this leads to different semantic interpretations in large-scale group decision-making (LSGDM). These interpretations evolve with changes in information, highlighting the necessity for personalized individual semantics (PISs) continuous learning. This study introduces a self-organized approach to achieving consensus in LSGDM through PISs continuous learning. First, we present a continuous learning model for PISs that employs a consistency-driven approach for updates. Next, a novel clustering algorithm is designed to partition large populations based on individual cognitive similarity, thereby reducing computational complexity. Subsequently, an adaptive feedback adjustment mechanism is introduced, integrating identification-direction rules and optimization-based consensus rules to enhance group consensus. Finally, the applicability of the proposed approach is demonstrated through a case study on HB-HTA for managing medical consumables.

关键词： Large-scale linguistic group decision-making Consensus reaching process Hospital-based health technology assessment continuous learning Personalised individual semantics

来源：评论

学校读者我要写书评

暂无评论

Secure and Efficient continuous learning Model for Traffic Flow Prediction

引用

IEEE TRANSACTIONS ON NETWORK AND SERVICE MANAGEMENT 2024年第4期21卷 4900-4911页

作者： Le, Junqing Zhang, Di Yang, Fan Xiang, Tao Liao, Xiaofeng Chongqing Univ Coll Comp Sci Chongqing 400044 Peoples R China Chongqing Univ Key Lab Dependable Serv Comp Cyber Phys Soc Minist Educ Chongqing 400044 Peoples R China

High-performance traffic flow prediction models provide reliable future road information and optimize traffic navigation systems. However, the traffic data used for model learning contains lots of private information, and the existing privacy-preserving strategies always reduce the accuracy of prediction models. Besides, an effective traffic flow prediction model needs to be continuously and rapidly updated to adapt to dynamic changes in the traffic environment. Thus, we propose a Secure and Efficient continuous learning Model (SE-CLM) based on broad learning, spatial correlation, and adaptive sampling processing techniques to realize accurate and efficient traffic flow prediction under strong privacy protection. Specifically, SE-CLM is constructed on the broad network architecture to enable fast and continuous model training. This model is trained on a cloud server by combining the spatial correlation of traffic flows, to achieve accurate traffic flow prediction. Besides, an adaptive sampling strategy is designed to further improve the prediction accuracy of the model under the protection with differential privacy (DP), where the budget allocation for DP is optimized by adaptively sampling traffic flows with different timestamps for noise perturbation processing. Furthermore, the experimental simulations are conducted in real vehicular mobility datasets. The experimental results show that the designed spatial-based SE-CLM achieve more accurate and efficient traffic flow prediction than those of the other existing schemes. The adaptive sampling strategy not only significantly reduces the DP-noise added in traffic flows but also a 20% reduction in communication volume compared to other strategies. Finally, the security analysis also verifies that SE-CLM satisfies w-event epsilon -DP.

关键词： Predictive models Adaptation models Data models Privacy Training Protection Roads Prediction model continuous learning security traffic flows cloud computing

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：