检索结果-内蒙古大学图书馆

41st International Conference on Machine Learning, ICML 2024

作者： Wang, Deng-Bao Zhang, Min-Ling School of Computer Science and Engineering Southeast University Nanjing China Key Lab. of Computer Network and Information Integration Southeast University MOE China

Although deep neural networks have achieved remarkable success, they often exhibit a significant deficiency in reliable uncertainty calibration. This paper focus on model calibratability, which assesses how amenable a model is to be well recalibrated post-hoc. We find that the widely used weight decay regularizer detrimentally affects model calibratability, subsequently leading to a decline in final calibration performance after post-hoc calibration. To identify the underlying causes leading to poor calibratability, we delve into the calibratability of intermediate features across the hidden layers. We observe a U-shaped trend in the calibratability of intermediate features from the bottom to the top layers, which indicates that over-compression of the top representation layers significantly hinders model calibratability. Based on the observations, this paper introduces a weak classifier hypothesis, i.e., given a weak classification head that has not been over-trained, the representation module can be better learned to produce more calibratable features. Consequently, we propose a progressively layer-peeled training (PLP) method to exploit this hypothesis, thereby enhancing model calibratability. Our comparative experiments show the effectiveness of our method, which improves model calibration and also yields competitive predictive performance. Copyright 2024 by the author(s)

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Leveraging Cross-Augmentation Consensus and Conflict for Semi-supervised Semantic Segmentation 27th

Leveraging Cross-Augmentation Consensus and Conflict for Se...

引用

27th International Conference on Pattern Recognition, ICPR 2024

作者： Cao, Junhao Chen, Junyi Huang, Sibo Zhang, Dongyu School of Computer Science and Engineering Sun Yat-sen University Guangzhou China Network and Information Center Huizhou University Huizhou China

ISBN: (纸本)9783031783975

Semi-supervised semantic segmentation leverages both labeled and unlabeled images to accomplish pixel-wise classification task. Within this field, the weak-to-strong consistency regularization has been widely popularized and has become a standard approach. However, unidirectional regularization often leads to the ignorance of correct but filtered predictions and brings the noise of wrong but confident predictions. To address these inherent flaws, we fully leverage Cross-Augmentation Consensus and Conflict (CACC), including Augmentation Feedback Mechanism (AFM) and Category Threshold Controller (CTC). AFM aims to mitigate the influence of incorrect predictions with high-confidence and mine unconfident but accurate predictions by re-weighting the pixel-wise pseudo supervision and applying supplementary regularization. Concurrently, CTC adopts category-specific thresholds by considering the model’s overall performance and the varying category-specific learning difficulty. Experimental results on benchmark datasets demonstrate the superior performance of our method, showcasing its effectiveness in improving semi-supervised semantic segmentation. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Semantic Segmentation

来源：评论

学校读者我要写书评

暂无评论

Generalization Analysis for Label-Specific Representation Learning 38

Generalization Analysis for Label-Specific Representation Le...

引用

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Zhang, Yi-Fan Zhang, Min-Ling School of Cyber Science and Engineering Southeast University Nanjing210096 China School of Computer Science and Engineering Southeast University Nanjing210096 China Key Laboratory of Computer Network and Information Integration Southeast University Ministry of Education China

Label-specific representation learning (LSRL), i.e., constructing the representation with specific discriminative properties for each class label, is an effective strategy to improve the performance of multi-label learning. However, the generalization analysis of LSRL is still in its infancy. The existing theory bounds for multilabel learning, which preserve the coupling among different components, are invalid for LSRL. In an attempt to overcome this challenge and make up for the gap in the generalization theory of LSRL, we develop a novel vector-contraction inequality and derive the generalization bound for general function class of LSRL with a weaker dependency on the number of labels than the state of the art. In addition, we derive generalization bounds for typical LSRL methods, and these theoretical results reveal the impact of different label-specific representations on generalization analysis. The mild bounds without strong assumptions explain the good generalization ability of LSRL. © 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

Rail Transit Event Extraction based on the Fusion of Semantic and Local Dependency Features 15

Rail Transit Event Extraction based on the Fusion of Semanti...

引用

15th IEEE International Conference on Knowledge Graph, ICKG 2024

作者： Chen, Qiang Meng, Haining Bao, Yihao Yang, Mingsong Tong, Xinyu Zhao, Qin Khan, Muhammad Bilal Hei, Xinhong Xi'an University of Technology School of Computer Science and Engineering Xi'an China Xi'an University of Technology School of Computer Science and Engineering Shaanxi Key Laboratory of Network Computing and Security Technology Xi'an China Xi'an University of Technology School of Civil Engineering and Architecture Xi'an China

ISBN: (纸本)9798331508821

Event extraction in rail transit plays a vital role in all stages of design, construction, and final acceptance. Event extraction is a key part of building event logic graph (ELG). However, the difficulty in obtaining semantic features and event local dependencies poses a challenge to accurate event extraction, which hinders the informatization and intelligent development of rail transit projects. To address these challenges, we proposes an event extraction method that fuses semantic and local dependency features. The study initially preprocesses the rail transit design codes dataset and performs sequence labeling. Then, the event trigger words are mined by combining the term frequency-inverse document frequency (TF-IDF) and minimum edit distance (MED) algorithms. Based on the semantic similarity measurement method and the trigger word mining results, the features representation of event semantic information is enhanced by using the Bidirectional Encoder Representations from Transformers (BERT) model and Bi-directional Long Short-Term Memory (BiLSTM) model. At the same time, event arguments are mined and a local dependency graph is generated based on their mining results, and then the enhanced local dependency features are obtained by using the graph convolutional network (GCN). Finally, the event semantic features and local dependency features are fused, and the conditional random field (CRF) is used to predict the labels to obtain the event extraction results. Through the comparison with five baseline methods on the datasets of rail transit design codes, it is found that the Precision, Recall, and F1 score of the proposed method are improved, with an average improvement of 9.604% on Precision, 9.342% on Recall, and 9.730% on F1 score, respectively. The contribution of the sub-modules to the overall performance is further evaluated through ablation experiments, which confirms the effectiveness and feasibility of the proposed method. © 2024 IEEE.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Joint Optimization of UAV Deployment and Task Computation Offloading Decision in UAV-assisted Edge Computing network 25

Joint Optimization of UAV Deployment and Task Computation Of...

引用

25th IEEE International Conferences on High Performance Computing and Communications, 9th International Conference on Data science and Systems, 21st IEEE International Conference on Smart City and 9th IEEE International Conference on Dependability in Sensor, Cloud and Big Data Systems and Applications, HPCC/DSS/SmartCity/DependSys 2023

作者： Liu, Yichuan Tu, Jinbin Wang, Yun School of Computer Science and Engineering Southeast University Key Lab of Computer Network and Information Integration Moe Nanjing China

ISBN: (纸本)9798350330014

The UAVs' deployment decision and task computation offloading decision in the UAV-assisted edge computing network significantly impact the operating efficiency of edge network. On the basis of this, the Optimization Model for UAV Cluster Deployment and Computation Offloading Decision (OMUCDCOD) is established. The model jointly optimizes the number, location of UAV s, and task computation offloading decision. Different from previous studies, this model regards the terminal devices in the edge network as virtual MEC servers, and introduces the collaborative computing mode. The task computation offloading decision made can effectively utilize the computing capabilities offered by the edge network. Considering that the two problems of UAV deployment decision and task computation offloading decision are intricately interconnected, we propose a two-layer optimization algorithm combining K-Means and ant colony algorithm (ToKmAc) to solve OMUCDCOD. ToKmAc is divided into upper and lower layers to solve this optimization problem. The upper layer uses K-Means to solve the UAV deployment decision, that is, the number and location of UAVS;the lower layer employs ant colony algorithm to solve computation offloading decision. Finally, extensive experiments verify the effectiveness of OMUCDCOD and ToKmAc. © 2023 IEEE.

关键词： Unmanned aerial vehicles (UAV)

来源：评论

学校读者我要写书评

暂无评论

Dynamic Visible Light Positioning Based on Enhanced Visual Target Tracking

引用

China Communications 2023年第10期20卷 276-291页

作者： Xiangyu Liu Jingyu Hao Lei Guo Song Song School of Information Science and Engineering Shenyang Ligong UniversityShenyang 110158China Department of Computer Science and Engineering Southern University of Science and TechnologyShenzhen 518055China School of Computer Science and Engineering Northeastern UniversityShenyang 110819China School of Communication and Information Engineering Chongqing University of Posts and TelecommunicationsChongqing 400065China Hangzhou Institute of Advanced Technology Hangzhou 310056China Institute of Intelligent Communications and Network Security Chongqing University of Posts and TelecommunicationsChongqing 400065China

In visible light positioning systems,some scholars have proposed target tracking algorithms to balance the relationship among positioning accuracy,real-time performance,and ***,there are still two problems:(1)When the captured LED disappears and the uncertain LED reappears,existing tracking algorithms may recognize the landmark in error;(2)The receiver is not always able to achieve positioning under various moving *** this paper,we propose an enhanced visual target tracking algorithm to solve the above ***,we design the lightweight recognition/demodulation mechanism,which combines Kalman filtering with simple image preprocessing to quickly track and accurately demodulate the ***,we use the Gaussian mixture model and the LED color feature to enable the system to achieve positioning,when the receiver is under various moving *** results show that our system can achieve high-precision dynamic positioning and improve the system’s comprehensive performance.

关键词： visible light positioning visual target tracking gaussian mixture model kalman filtering system performance

来源：评论

学校读者我要写书评

暂无评论

T-Sorokin: A General Mobility Model in Opportunistic networks 26

T-Sorokin: A General Mobility Model in Opportunistic Network...

引用

26th International Conference on computer Supported Cooperative Work in Design, CSCWD 2023

作者： Tu, Jinbin Li, Qing Wang, Yun Southeast University Key Lab of Computer Network and Information Integration Moe School of Computer Science and Engineering Nanjing China

ISBN: (纸本)9798350331684

The opportunistic networks are a kind of ad hoc networks that rely on the chance of nodes meeting to transmit messages. Acting as an effective supplement to 4G and 5G networks in some special scenarios where hardware devices are limited, the opportunistic networks have a significant application in health monitoring, warning broadcasting, disaster relief, and so on. The mobility model is one of the research focuses on the opportunistic networks. On the basis of the social mobility theory proposed by Sorokin, a general mobility model, which is suited for various scenarios, called T-Sorokin is proposed. This model is described as a seven-tuple and implemented on the Opportunistic network Environment simulator and fits both Infocom06 and Rome taxi data set, which includes different areas ranging from hotel to city and different mobile units ranging from person to taxi. The results of experiments demonstrate that the T-Sorokin model has the advantage of generality, simplicity, and accuracy. It can simply establish movement tracks close to real data under different scenarios. © 2023 IEEE.

关键词： 5G mobile communication systems

来源：评论

学校读者我要写书评

暂无评论

Endogenous Security of FQn networks: Adaptive System-Level Fault Self-Diagnosis

引用

IEEE Transactions on Reliability 2024年第3期73卷 1659-1668页

作者： Lin, Yuhang Lin, Limei Huang, Yanze Xu, Li Hsieh, Sun-Yuan Fujian Normal University College of Computer and Cyber Security Fujian Provincial Key Laboratory of Network Security and Cryptology Fuzhou350117 China Fujian University of Technology School of Computer Science and Mathematics Fuzhou350118 China National Cheng Kung University Department of Computer Science and Information Engineering Tainan701 Taiwan

Endogenous security has the ability to discover, eliminate, and solve internal security problems and hidden dangers within the network, and is a superior technology to ensure future network security. The t/k-diagnosis strategy, as a strong and adaptive self-diagnosis strategy, is an important part for ensuring endogenous security. Moreover, the folded hypercube (FQn) as a data transmission network (e.g., optical network) topology offers new potential for the construction of large-scale, high data throughput, and low-latency systems, such as computing network, human-cyber-physical systems, and smart grid. However, there are few studies on endogenous security based on FQn networks. Therefore, this article designs an adaptive system-level fault self-diagnosis strategy, namely Fast t/k-Diagnosis Under Maeng-Malek Model (Ftk-DIAG-MM*) to diagnosis the faulty vertices in FQn network under the Maeng-Malek model (MM∗ mod). Then, we provide a proof of the algorithm correctness theoretically by the fault tolerance of FQn network. It is derived by theoretical derivation that the t/k-diagnosability inherent to the FQn network is (n+1)(k+1)-k(k+3)/2. The simulation experiments demonstrate that the designed Ftk-DIAG-MM∗ strategy can correctly diagnose all vertices within the range allowed by the diagnosability, and still has a great performance when it exceeds the range allowed by the diagnosability. It greatly enhances the fault diagnosis capability of FQn network in the circumstance of misdiagnosing a few vertices, which provides an important theoretical basis for the reliability and endogenous security of FQn networks. © 1963-2012 IEEE.

关键词： Topology

来源：评论

学校读者我要写书评

暂无评论

Improving Speech Enhancement by Cross- and Sub-band Processing with State Space Model

Improving Speech Enhancement by Cross- and Sub-band Processi...

引用

2025 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2025

作者： Li, Jizhen Tu, Weiping Yang, Yuhong Xu, Xinmeng Zhang, Yiqun Ren, Yanzhen NERCMS School of Computer Science Hubei Luojia Laboratory Wuhan University China Hubei Key Laboratory of Multimedia and Network Communication Engineering Wuhan University China School of Cyber Science and Engineering Wuhan University China

ISBN: (纸本)9798350368741

Recently, the state space model (SSM) represented by Mamba has shown remarkable performance in long-term sequence modeling tasks, including speech enhancement. However, due to substantial differences in sub-band features, applying the same SSM to all sub-bands limits its inference capability. Additionally, when processing each time frame of the time-frequency representation, the SSM may forget certain high-frequency information of low energy, making the restoration of structure in the high-frequency bands challenging. For this reason, we propose Cross- and Sub-band Mamba (CSMamba). To assist the SSM in handling different sub-band features flexibly, we propose a band split block that splits the full-band into four sub-bands with different widths based on their information similarity. We then allocate independent weights to each subband, thereby reducing the inference burden on the SSM. Furthermore, to mitigate the forgetting of low-energy information in the high-frequency bands by the SSM, we introduce a spectrum restoration block that enhances the representation of the cross-band features from multiple perspectives. Experimental results on the DNS Challenge 2021 dataset demonstrate that CSMamba outperforms several state-of-the-art (SOTA) speech enhancement methods in three objective evaluation metrics with fewer parameters. © 2025 IEEE.

关键词： band split spectrum restoration speech enhancement state space model

来源：评论

学校读者我要写书评

暂无评论

Learning group interaction for sports video understanding from a perspective of athlete

引用

Frontiers of computer science 2024年第4期18卷 175-188页

作者： Rui HE Zehua FU Qingjie LIU Yunhong WANG Xunxun CHEN Intelligent Recognition and Image Processing(IRIP)Lab School of Computer Science and EngineeringBeihang UniversityBeijing 100191China Hangzhou Innovation Institute Behang UniversityHangzhou 310051China National Computer Network Emergency Response Technical Team/Coordination Center of China(CNCERT or CNCERT/CC) Beijing 100029China

Learning activities interactions between small groups is a key step in understanding team sports *** research focusing on team sports videos can be strictly regarded from the perspective of the audience rather than the *** team sports videos such as volleyball and basketball videos,there are plenty of intra-team and inter-team *** this paper,a new task named Group Scene Graph Generation is introduced to better understand intra-team relations and inter-team relations in sports *** tackle this problem,a novel Hierarchical Relation network is *** all players in a video are finely divided into two teams,the feature of the two teams’activities and interactions will be enhanced by Graph Convolutional networks,which are finally recognized to generate Group Scene *** evaluation,built on Volleyball dataset with additional 9660 team activity labels,a Volleyball+dataset is proposed.A baseline is set for better comparison and our experimental results demonstrate the effectiveness of our ***,the idea of our method can be directly utilized in another video-based task,Group Activity *** show the priority of our method and display the link between the two ***,from the athlete’s view,we elaborately present an interpretation that shows how to utilize Group Scene Graph to analyze teams’activities and provide professional gaming suggestions.

关键词： group scene graph group activity recognition scene graph generation graph convolutional network sports video understanding

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：