检索结果-内蒙古大学图书馆

Journal of Applied and Numerical Optimization 2023年第2期5卷 237-253页

作者： Tong, Kai Ke, Xiao Fujian Provincial Key Laboratory of Networking Computing and Intelligent Information Processing College of Computer and Data Science Fuzhou University Fuzhou350116 China Key Laboratory of Spatial Data Mining & Information Sharing Ministry of Education Fuzhou350003 China

Deep neural networks require correct label annotation during supervised learning. It is inevitable, however, that some labels are noisy during the labeling process. A deep neural network retains incorrect labels during training, resulting in a degradation of performance. Therefore, it is essential to identify samples with potentially correct labels. In state-of-the-art methods, small-loss samples are chosen for subsequent training through a sample selection strategy. Howerver, it typically ignores the imbalance in noise ratios between mini-batches when performing sample selection within each minibatch. Further, numerous valuable samples with high losses are discarded, which adversely affects the generalization performance of the model, particularly under conditions of high noise ratios. To this end, this paper proposes IdentifyMix, an effective two-stage learning approach for noisy robust learning that combines an unique sample selection strategy and the semi-supervised learning technique. By observing how the dynamics of network training are changing, AUM (Area Under the Margin) provides a criterion that is applied in this research to identify mislabeled data. Moreover, by combining semi-supervised learning with contrastive learning and data augmentation, it is possible to extract more useful information from mislabeled samples. Experiments on several synthetic and real-world noise benchmarks demonstrate the effectiveness of IdentifyMix compared with state-of-the-art methods. © 2023 Journal of Applied and Numerical Optimization.

关键词： Image classification

来源：评论

学校读者我要写书评

暂无评论

Accelerating Unsupervised Federated Graph Neural Networks via Semi-asynchronous Communication 18th

Accelerating Unsupervised Federated Graph Neural Networks vi...

引用

18th CCF Conference on Computer Supported Cooperative Work and Social computing, ChineseCSCW 2023

作者： Liao, Yuanming Wu, Duanji Lin, Pengyu Guo, Kun College of Computer and Data Science Fuzhou University Fuzhou350108 China Fujian Key Laboratory of Network Computing and Intelligent Information Processing Fuzhou University Fuzhou350108 China Key Laboratory of Spatial Data Mining and Information Sharing Ministry of Education Fuzhou350108 China

ISBN: (纸本)9789819996360

Graph neural networks have shown excellent performance in many fields owing to their powerful processing ability of graph data. In recent years, federated graph neural network has become a reasonable solution due to the enactment of privacy-related regulations. However, frequent communication between the coordinator and participants in federated graph neural network results in longer model training time and consumes many communication resources. To address this challenge, in this paper, we propose a novel semi-asynchronous federated graph learning communication protocol that simultaneously alleviates the negative impact of stragglers(slow participants) and accelerate the training process in the unsupervised federated graph neural network scenario. First, the weighted enforced synchronization strategy is intended to preserve the information carried by stragglers while preventing their stale models from harming the global model update. Second, the adaptive local update strategy is developed to make the local model of the participant with poor computing performance as close as possible to the global model. Experiments combine federated learning with graph contrastive learning. The results demonstrate that our proposed protocol outperforms the existing protocols in real-world networks. © 2024, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Internet protocols

来源：评论

学校读者我要写书评

暂无评论

Generalized Self-Adaption Network for Domain Adaptation 2

Generalized Self-Adaption Network for Domain Adaptation

引用

2nd International Conference on Artificial Intelligence and intelligent Information Processing, AIIIP 2023

作者： Bian, Yongheng Ke, Xiao College of Computer and Data Science Fuzhou University Fujian Provincial Key Laboratory of Networking Computing and Intelligent Information Processing Fuzhou350116 China Key Laboratory of Spatial Data Mining & Information Sharing Ministry of Education Fuzhou350003 China

ISBN: (纸本)9798350371451

The purpose of domain adaptation is to transfer knowledge learned in the labeled source domain to unlabeled but related target domains without requiring a large number of target domain labels. The latest method of domain adaptation mainly adjusts the distribution of the source and target domains to adapt to the target domain. Despite significant success, when there are significant distribution differences between the source and target domains, the adaptability in the target domain still decreases, which may be due to insufficient exploration of domain specific features leading to domain offset issues. To address this issue, we propose the GSA-NET framework in this article. We designed an adversarial dropout method to improve the generalization performance of the model in the target domain by pushing the decision boundary away from the feature dense region. At the same time, in order to effectively match the distribution of advanced features across domains, we deploy feature adaptation modules before the classification layer, which can significantly reduce domain differences. Our experiments on two cross domain benchmarks, Office Home and Office-31, have shown that the proposed method is significantly superior to existing methods. © 2023 IEEE.

关键词： Computer vision

来源：评论

学校读者我要写书评

暂无评论

ABCD-HN: An Artificial Network Benchmark for Community Detection on Heterogeneous Networks 18th

ABCD-HN: An Artificial Network Benchmark for Community Dete...

引用

18th CCF Conference on Computer Supported Cooperative Work and Social computing, ChineseCSCW 2023

作者： Liu, Junjie Guo, Kun Wu, Ling Fujian Provincial Key Laboratory of Network Computing and Intelligent Information Processing Fuzhou University Fuzhou350108 China College of Computer and Data Science/College of Software Fuzhou University Fuzhou350108 China Key Laboratory of Spatial Data Mining and Information Sharing Ministry of Education Fuzhou350108 China

ISBN: (纸本)9789819996360

Community detection is essential for identifying cohesive groups in complex networks. Artificial benchmarks are critical for evaluating community detection algorithms, offering controlled environments with known community structures. However, existing benchmarks mainly focus on homogeneous networks and overlook the unique characteristics of heterogeneous networks. This paper proposes a novel artificial benchmark, called ABCD-HN (Artificial Network Benchmark for Community Detection on Heterogeneous Networks), for community detection in heterogeneous networks. This benchmark enables the generation of artificial heterogeneous networks with controllable community quantity, node quantity, and community complexity. Additionally, an evaluation framework for artificial heterogeneous networks is proposed to assess their effectiveness. Experimental results demonstrate the effectiveness and usability of ABCD-HN as a benchmark for artificial heterogeneous networks. © 2024, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Heterogeneous networks

来源：评论

学校读者我要写书评

暂无评论

Few-Shot Object Detection Based on Generalized Features 2

Few-Shot Object Detection Based on Generalized Features

引用

2nd International Conference on Artificial Intelligence and intelligent Information Processing, AIIIP 2023

作者： Chen, Qiuqin Ke, Xiao College of Computer and Data Science Fuzhou University Fujian Provincial Key Laboratory of Networking Computing and Intelligent Information Processing Fuzhou350116 China Key Laboratory of Spatial Data Mining & Information Sharing Ministry of Education Fuzhou Fuzhou350003 China

ISBN: (纸本)9798350371451

Few-shot object detection aims to rapidly detect novel classes of objects using a minimal number of annotated instances. Compared to methods such as meta-learning, few-shot object detection based on transfer learning can achieve better performance. However, when training on base-class datasets, the model gradually improves its detection performance on base classes but does not enhance its generalization. A base-class model lacking generalization tends to result in lower detection performance for novel classes during fine-tuning. To address the aforementioned issues, this paper tackles the problem from two aspects: firstly, we introduce the Confusion-Dropblock Module (CDM) to perturb the model's features, thus enhancing feature generalization. Secondly, we incorporate Deformable Convolution Modules to mitigate the impact of object shapes and sizes on detection performance, thereby improving overall detection performance. Experimental results on the COCO dataset demonstrate the effectiveness of the proposed algorithm, outperforming the compared methods. © 2023 IEEE.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

Community Evolution Tracking Based on Core Node Extension and Edge Variation Discerning 17th

Community Evolution Tracking Based on Core Node Extension a...

引用

17th CCF Conference on Computer Supported Cooperative Work and Social computing, ChineseCSCW 2022

作者： Zhuang, Qifeng Yu, Zhiyong Guo, Kun College of Computer and Data Science Fuzhou University Fuzhou350108 China Fujian Provincial Key Laboratory of Network Computing and Intelligent Information Processing Fuzhou350108 China Key Laboratory of Spatial Data Mining and Information Sharing Ministry of Education Fuzhou350108 China

ISBN: (纸本)9789819923557

Communities exist anywhere in various complex networks, and community evolution tracking is one of the most well-liked areas of inquiry in the study of dynamic complex networks. Community evolution tracking has many applications in daily life, such as predicting social network behaviors or analyzing the spread of infectious illnesses. However, the majority of existing evolution tracking algorithms obtain community detection results before matching the community in tracking evolution events, making it difficult to trace the whole evolution of communities because of community matching errors. In addition, the majority of evolution tracking algorithms do not adequately account for the potential scenarios in community evolution, resulting in erroneous detection of evolution events. In this research, a community evolution tracking algorithm based on edge variation discerning and core node extension is proposed. First, we detect communities based on the core node extension strategy, which avoids the problem of community matching errors. Second, we track community evolution based on the edge variation discerning strategy, which fully considers various situations that may occur during the community evolution process. According to the outputs of our experiments, our system can effectively track the evolution of communities in synthetic dynamic networks. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Complex networks

来源：评论

学校读者我要写书评

暂无评论

AUCSeg: AUC-oriented Pixel-level Long-tail Semantic Segmentation 38

AUCSeg: AUC-oriented Pixel-level Long-tail Semantic Segmenta...

引用

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Han, Boyu Xu, Qianqian Yang, Zhiyong Bao, Shilong Wen, Peisong Jiang, Yangbangyan Huang, Qingming Key Lab. of Intelligent Information Processing Institute of Computing Technology CAS China School of Computer Science and Tech. University of Chinese Academy of Sciences China Peng Cheng Laboratory China Key Laboratory of Big Data Mining and Knowledge Management CAS China

The Area Under the ROC Curve (AUC) is a well-known metric for evaluating instance-level long-tail learning problems. In the past two decades, many AUC optimization methods have been proposed to improve model performance under long-tail distributions. In this paper, we explore AUC optimization methods in the context of pixel-level long-tail semantic segmentation, a much more complicated scenario. This task introduces two major challenges for AUC optimization techniques. On one hand, AUC optimization in a pixel-level task involves complex coupling across loss terms, with structured inner-image and pairwise inter-image dependencies, complicating theoretical analysis. On the other hand, we find that mini-batch estimation of AUC loss in this case requires a larger batch size, resulting in an unaffordable space complexity. To address these issues, we develop a pixel-level AUC loss function and conduct a dependency-graph-based theoretical analysis of the algorithm's generalization ability. Additionally, we design a Tail-Classes Memory Bank (T-Memory Bank) to manage the significant memory demand. Finally, comprehensive experiments across various benchmarks confirm the effectiveness of our proposed AUCSeg method. The code is available at https://***/boyuh/AUCSeg. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Unified Stream and Batch Graph computing Model for Community Detection 17th

A Unified Stream and Batch Graph Computing Model for Commu...

引用

17th CCF Conference on Computer Supported Cooperative Work and Social computing, ChineseCSCW 2022

作者： Dai, Jinkun Wu, Ling Guo, Kun Fujian Provincial Key Laboratory of Network Computing and Intelligent Information Processing Fuzhou University Fuzhou350108 China College of Computer and Data Science Fuzhou University Fuzhou350108 China Key Laboratory of Spatial Data Mining and Information Sharing Ministry of Education Fuzhou350108 China

ISBN: (纸本)9789819923557

An essential challenge in graph data analysis and mining is to simply and effectively deal with large-scale network data that is expanding dynamically. Although batch-based parallel graph computation frameworks have better accuracy, they cannot process incremental data on time and need to be recomputed. To process real-time data, stream processing applications need to be redeveloped, which increases the redundancy of work, and some existing dynamic graph computation schemes are not generalizable. This paper proposes a unified stream and batch graph computing model(USBGM). The model is compatible with both stream and batch graph computing. Graph operators and algorithms developed based on the model can handle stream and batch graph data in a unified manner. The experiments on real-world and artificial networks verified the effectiveness and efficiency of the model. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Population dynamics

来源：评论

学校读者我要写书评

暂无评论

Globally Consistent Vertical Federated Graph Autoencoder for Privacy-Preserving Community Detection 17th

Globally Consistent Vertical Federated Graph Autoencoder for...

引用

17th CCF Conference on Computer Supported Cooperative Work and Social computing, ChineseCSCW 2022

作者： Fang, Yutong Huang, Qingqing Ye, Enjie Guo, Wenzhong Guo, Kun Chen, Xiaoqi Fujian Provincial Key Laboratory of Network Computing and Intelligent Information Processing Fuzhou University Fuzhou350108 China College of Computer and Data Science/College of Software Fuzhou University Fuzhou350108 China Key Laboratory of Spatial Data Mining and Information Sharing Ministry of Education Fuzhou350108 China

ISBN: (纸本)9789819923557

Community detection is a trendy area in research on complex network analysis and has a wide range of real-world applications, like advertising. As people become increasingly concerned about privacy, protecting participants’ privacy in distributed community detection has become a new challenge. When applied to not identically and independently distributed (Non-IID) data, most federated graph algorithms suffer from the weight divergence problem caused by diversified training of local and global models, resulting in accuracy degradation. Furthermore, the privacy protection approaches based on anonymization, such as differential privacy (DP), and cryptography, such as homomorphic encryption (HE), incur accuracy loss and high time consumption, respectively. In this paper, we propose a globally consistent vertical federated graph autoencoder (GCVFGAE) algorithm, which builds a globally consistent model among the coordinator and all participants to solve the Non-IID graph data problem. As well, an attribute blinding strategy based on security aggregation is developed to protect the network privacy of each participant without losing accuracy. Both real-world and artificial networks’ experiments show that our algorithm reaches higher accuracy than the existing vertical federated graph neural networks (GNNs) and the simple distributed graph autoencoder without federated learning and detects communities identical to those found by the standard graph autoencoder (GAE). © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Population dynamics

来源：评论

学校读者我要写书评

暂无评论

Horizontal Federated Traffic Speed Prediction Base on Secure Node Attribute Aggregation 17th

Horizontal Federated Traffic Speed Prediction Base on Secur...

引用

17th CCF Conference on Computer Supported Cooperative Work and Social computing, ChineseCSCW 2022

作者： Ye, Enjie Guo, Kun Guo, Wenzhong Chen, Dangrun Zhang, Zihan Li, Fuan Zheng, JiaChen Fujian Provincial Key Laboratory of Network Computing and Intelligent Information Processing Fuzhou University Fuzhou350108 China College of Computer and Data Science/College of Software Fuzhou University Fuzhou350108 China Key Laboratory of Spatial Data Mining and Information Sharing Ministry of Education Fuzhou350108 China

ISBN: (纸本)9789819923847

Federated graph learning has been widely used in distributed graph machine learning tasks. The data distribution of existing graph-based federated Spatio-temporal prediction methods is mainly segmented by graph topology. However, in the real-world Spatio-temporal traffic speed prediction task, a location will have data from different devices belonging to different companies. A node may have multi-party information in the real-world distributed traffic speed prediction scenario. The difference in multi-party information leads to the information not being fully utilized. Moreover, the direct transmission of node embedding in the federated learning process may also risk privacy leaks. Using homomorphic encryption and other encryption methods will bring a high computational overhead. Therefore we propose a new distributed privacy-preserving traffic speed prediction algorithm, which uses secure node attribute aggregation strategy(SNAAS) to apply to the multi-party collaborative traffic speed prediction scenario when the graph topology structure is public. At the same time, secret sharing technology is used in SNAAS to protect the attribute matrix and reduce the overhead of secret computing. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Forecasting

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：