检索结果-内蒙古大学图书馆

SSRN 2024年

作者： Tang, LingXiang Qin, XiaoXiao Liu, YongChao Huo, BoFeng School of Mathematics and Statists Qinghai Normal University China The State Key Laboratory of Tibetan Intelligent Information Processing and Application China Academy of Plateau Science and Sustainability Xining Qinghai810000 China

In 1941, Brooks states that for a graph $G$ with $\Delta(G)\geq 3$, $\chi(G)\leq max\{\Delta(G), \omega(G)\}$. In 1977, Borodin and Kostochka conjectured a similar result that if $\Delta(G)\geq 9$,then $\chi(G)\leq max\{\Delta(G)-1, \omega(G)\}$. The conjecture is still a open *** this paper, we show that the conjecture is true for $\{P_4 \cup P_1, diamond, house\}$-freegraphs and $\{P_4 \cup P_1, cochair,bull\}$-free graphs. © 2024, The Authors. All rights reserved.

关键词： Graphic methods

来源：评论

学校读者我要写书评

暂无评论

Remote Sensing Image Captioning Using Hire-MLP 4

Remote Sensing Image Captioning Using Hire-MLP

引用

4th International Conference on Computer Vision, Image and Deep Learning, CVIDL 2023

作者： Wei, Yichao Li, Lin Geng, Shengling School of Computer Science Qinghai Normal University Xining China The State Key Laboratory of Tibetan Intelligent Information Processing and Application Xining China Qinghai Normal University Academy of Plateau Science and Sustainability Xining China National Qinghai-Tibet Plateau Scientific Data Center Qinghai Sub-center Xining China

ISBN: (纸本)9798350326444

Remote sensing image captioning has been widely applied to traffic management, geographic research, etc. Although the neural network approach has been successfully improving the performance of the Remote sensing image captioning system, Remote sensing image captioning is still facing object identification challenges due to the small size of objects, uneven distribution, and high coupling with the surrounding image background. In this paper, we propose a novel remote sensing image captioning encoder-decoder model Hierarchical rearrangement-Multi-Layer Perceptron (HMLP) whose encoder adapts hierarchical rearrangement-multi-layer perceptron to improve the capability of objection recognition. Extensive experiments have been conducted to testify HMLP by three datasets RSCID, UCM-caption, and NWPU-caption. Results show that HMLP outperforms many image captioning systems in the evaluation metrics BLEU4, METEOR, ROUGE-L, and CIDEr. © 2023 IEEE.

关键词： Computer vision

来源：评论

学校读者我要写书评

暂无评论

PT-MVSNet: Overlapping Attention Multi-view Stereo Network with Transformers 4

PT-MVSNet: Overlapping Attention Multi-view Stereo Network w...

引用

4th International Conference on Computer Vision, Image and Deep Learning, CVIDL 2023

作者： Zhang, Song Li, Lin Qiao, Jiangxuan School of Computer Science Qinghai Normal University Xining China The State Key Laboratory of Tibetan Intelligent Information Processing and Application Xining China Qinghai Normal University Academy of Plateau Science and Sustainability Xining China National Qinghai-Tibet Plateau Scientific Data Center Qinghai Sub-center Xining China

ISBN: (纸本)9798350326444

In this paper, we propose a new multi-view stereo vision model PT -MVSNet based on multi-view stereo (MVS). Multi-view stereo is a successful reconstruction method that uses multiple images to reconstruct a 3D scene. It has been applied in many practical scenes such as architecture, cultural heritage protection, and map making. MVS still faces a lot of challenges, including inaccurate feature matching, excessive image noise, and overly complex computation. To solve the feature-matching inaccuracy problem, we take the Transformer model as the main structure in the feature-matching and add a patch-based overlap attention module (POLA). In this paper, we proposed PT-MVSNet can solve the image feature extraction problem more effectively. To validate the effectiveness of the model, we conducted experiments on the DTU dataset and evaluated its performance by two evaluation metrics. The experiment results show that our method outperforms the latest methods, whose accuracy and completeness reach 0.386 and 0.271 respectively. © 2023 IEEE.

关键词： Feature extraction

来源：评论

学校读者我要写书评

暂无评论

A semantic segmentation network based on CNN and Swin transformer for landslide detection 6

A semantic segmentation network based on CNN and Swin transf...

引用

2024 6th International Conference on Artificial Intelligence Technologies and applications, ICAITA 2024

作者： Dai, Xiangyan Li, Lin School of Computer Science Qinghai Normal University Qinghai Xining810008 China The State Key Laboratory of Tibetan Intelligent Information Processing and Application Qinghai Xining810016 China Academy of Plateau Science and Sustainability Qinghai Normal University Qinghai Xining810008 China Qinghai Sub-Center National Qinghai-Tibet Plateau Scientific Data Center Qinghai Xining810008 China

To address the issues of poor object localization, difficulty in object recognition, and inadequate segmentation performance of CNN-based semantic segmentation methods in landslide detection, a semantic segmentation model based on the Swin transformer and ResUNet architectures for landslide detection in remote sensing images is proposed in this paper. This method combines the global feature capabilities of the Swin transformer with the local feature extraction abilities of CNN and integrates our proposed RCCT module for landslide segmentation in remote sensing images. We apply this model to the identification and extraction of landslide disasters in Menyuan County and the Sanjiangyuan region of Qinghai Province, China, achieving a promising F1 Score of 81.91. Furthermore, we conducted a comprehensive evaluation of our proposed model on the BiJie Landslide dataset in comparison to four recently developed methods for landslide recognition. The experimental findings indicate the superior performance and effectiveness of our model over the comparative approaches. © 2024 Institute of Physics Publishing. All rights reserved.

关键词： Semantic Segmentation

来源：评论

学校读者我要写书评

暂无评论

End-to-End tibetan Speech Recognition Study Based on Byte Pair Coding 23

引用

23rd Chinese National Conference on Computational Linguistics, CCL 2024

作者： Cai, Yuqing Wang, Chao Duojie, Renzeng Zhu, Yulei Zhang, Jin Tashi, Nyima Tibet University School of Information Science and Technology Lhasa850000 China State Key Laboratory of Tibetan Intelligent Information Processing and Application Lhasa850000 China Qinghai Normal University College of Computer Xining810016 China

ISBN: (纸本)9780000000002

Aiming at the problems of inconsistent modeling units and unsatisfactory recognition results in tibetan end-to-end speech recognition research, this paper proposes a BPE-Conformer-CTC/Attention end-to-end tibetan speech recognition method. Firstly, the method adopts the byte-pair encoding algorithm for speech modeling, which solves the problem of inconsistent modeling units in tibetan speech recognition by repeatedly merging the most frequent character pairs, segmenting the text into manageable and meaningful units, and balancing the granularity of modeling units. Secondly, the Con-former encoder is utilized to effectively fuse the global and local dependencies of audio sequences, thus enhancing the model's representational capability. Finally, the alignment and decoding process is accelerated by a joint CTC/Attention decoding strategy, which in turn improves the accuracy and efficiency of the recognition results. Experimental results on the open-source datasets XBMU-AMD031 and TIBMD@MUC show that the proposed BPE-Conformer-CTC/Attention model achieves a word error rate of 9.0% and 4.6%, respectively, and compared to the baseline Transformer-CTC/Attention model, the word error rate is reduced by 14.2% and 30.3%, respectively. This proposed approach provides an effective solution for the tibetan end-to-end speech recognition task. © 2024 China National Conference on Computational Linguistics.

关键词： Decoding

来源：评论

学校读者我要写书评

暂无评论

Cascading failure analysis of uniform double-layer hyper-networks based on the Couple Map Lattice model

Cascading failure analysis of uniform double-layer hyper-net...

引用

IEEE International Conference on High Performance Computing and Communications (HPCC)

作者： Shujie Gao Xiujuan Ma Fuxiang Ma Bin Zhou The State Key Laboratory of Tibetan Intelligent Information Processing and Application School of Computer Qinghai Normal University Xining China

Cascading failures are a class of dynamical behaviors that are widely present in various types of complex systems. The study of cascading failures has also been a hot topic for researchers. The results have shown that the process of cascading failures varies greatly depending on the structure of the network. Based on the double-layer hyper-network structure, considering the types of hyper-networks and the way of inter-layer connection, we construct four types of 3 uniform double-layer hyper-networks. In addition, the Couple Map Lattice (CML) cascading failure model is proposed which is suitable for describing the cascading failure behavior of a double-layer hyper-network. By this model, we analyzed the cascading failure processes happen in hyper-edges of a uniform double-layer hyper-network. Furthermore, through theoretical analysis, an approximate solution for the external perturbation threshold is obtained, which makes the double-layer hyper-network reaches the global collapse. The results of this paper provide an important reference for optimizing the robustness of realistic double-layer hyper-networks. In practical applications, the robustness of the double-layer hyper-networks can be improved by reducing the inter-layer connectivity probability $p12$ and the coupling strength $\varepsilon$ between hyper-edges.

关键词： Couplings Analytical models Correlation Perturbation methods Power system protection Lattices Failure analysis

来源：评论

学校读者我要写书评

暂无评论

A practical adversarial attack on Graph Neural Networks by attacking single node structure

A practical adversarial attack on Graph Neural Networks by a...

引用

IEEE International Conference on High Performance Computing and Communications (HPCC)

作者： Yang Chen Zhonglin Ye Haixing Zhao Lei Meng Zhaoyang Wang Yanlin Yang The State Key Laboratory of Tibetan Intelligent Information Processing and Application School of Computer Science Qinghai Normal University Qinghai China

In recent years, Graph Neural Networks (GNNs) have performed significantly in node classification tasks. Re-search on GNNs adversarial attacks has recently attracted attention. Many works have demonstrated that single node attacks can fool GNNs in extreme scenarios. Existing single node attacks can be divided into two main types: 1) attacking the node features; 2) injecting a single fake node. However, they have several major drawbacks: 1) they set the attack node features, which makes them not efficient; 2) they require higher privileges and don't have strict budget constraints, which makes them less concealable and easier to notice. In this paper, we propose an attack with lower privileges and strict budget constraints -Single Node Structure Attack (SNSA). Specifically, SNSA is set more efficient structure attacks and does not require higher privileges to change links. Compared to other attack models, the attack space of SNSA is small and not easily detected by GNNs. Furthermore, we study the influence of the attack node selection on SNSA and observe that it is optimal to use a gradient to select an attack node. Experiments on three common datasets demonstrate the effectiveness of our model.

关键词： Degradation Graph neural networks Task analysis Periodic structures

来源：评论

学校读者我要写书评

暂无评论

Long Time Series Prediction of Meteorological Data Based on Improved Autoformer Model 2

Long Time Series Prediction of Meteorological Data Based on ...

引用

2nd International Conference on Artificial Intelligence and intelligent information processing, AIIIP 2023

作者： Song, Suxin Geng, Shengling Wang, Yuxiang School of Computer Science Qinghai Normal University Hutai Qinghai Xining810008 China Academy of Plateau Science and Sustainability People's Government of Qinghai Province Beijing Normal University Qinghai Haihu Xining 810004 China Qinghai Normal University Hutai The State Key Laboratory of Tibetan Intelligent Information Processing and Application Qinghai Xining810008 China

ISBN: (纸本)9798350371451

Aiming at the problem of low prediction accuracy caused by many uncertain factors in the prediction of long-time series of meteorological data, a long-time series prediction model of meteorological data based on improved Autoformer model (Wdformer) is proposed. Change the routine of sequence decomposition pretreatment and reduce the noise and irregularity in the data by introducing dynamic diffusion to simulate the changes of meteorological elements in the atmosphere;The wavelet transform time series decomposition module is designed to improve the gradual decomposition ability of the model for long time series. The experimental results of meteorological data prediction in a certain area show that the accuracy of the proposed Wdformer model is higher than that of the standard Autoformer model. © 2023 IEEE.

关键词： Wavelet decomposition

来源：评论

学校读者我要写书评

暂无评论

Three-Channel Graph Convolutional Network

Three-Channel Graph Convolutional Network

引用

IEEE International Conference on High Performance Computing and Communications (HPCC)

作者： Lei Meng Zhonglin Ye Haixing Zhao Yanlin Yang Yang Chen Zhaoyang Wang The State Key Laboratory of Tibetan Intelligent Information Processing and Application College of Computer Qinghai Normal University Tibetan Information Processing Engineering Technology and Research Center of Qinghai Province Xining China

At present, as an emerging direction of artificial intelligence, graph neural networks (GNNs) have developed rapidly and can solve many graph data problems efficiently. The real world is composed of a large number of networks. Compared with the widely studied single-relationship feature information networks, the real-world networks are mostly composed of multi-relationship feature information networks in which different types of objects are interconnected through complex relationships. Complex structural information and rich semantic information are included by multi-relationship feature information networks, thereby the real network can be restored more realistically. However, one type of relational feature of the network is merely considered by most of the existing GNNs, and the information of multi-relationship features and the hidden information between multi-relationship features are ignored in the network. In order to deal with these restrictions, a novel graph convolutional network (TCGCN) based on an attention mechanism is proposed, which aims to extract different relational features in multi-relationship feature networks and the hidden information between them more comprehensively, thereby improving the performance of GNNs in downstream tasks. In this paper, an improved attention mechanism is used to aggregate different types of single relational features, and the aggregated multiple single relational feature network information is fused with the mixed relational feature network information into the final node representation. The proposed TCGCN outperforms most other methods in extensive experiments conducted on three real-world datasets.

关键词： Smart cities Aggregates Semantics Feature extraction Graph neural networks Convolutional neural networks Data mining

来源：评论

学校读者我要写书评

暂无评论

FEMD: Feature Enhancement-aided Multimodal Feature Fusion Approach for Smart Contract Vulnerability Detection 30

FEMD: Feature Enhancement-aided Multimodal Feature Fusion Ap...

引用

30th IEEE International Conference on Parallel and Distributed Systems, ICPADS 2024

作者： Li, Juan Li, Rui Lu, Youshui Cai, Bowen Cao, Yulin Lin, Hang Li, Xingxing Li, Chan Qinghai Normal University School of Computer Science Xining China Xidian University The State Key Laboratory of Tibetan Intelligent Information Processing and Application School of Computer Science Qinghai Normal University School of Computer Science and Technology Xian China School of the Instrument Science and Technology School of Electrical Engineering Xi'an Jiaotong University Xian China School of Science and Engineering University of Minnesota - Twin City Minneapolis United States

ISBN: (纸本)9798331515966

Smart contracts, due to their immutability and transparency upon deployment, entail significant economic and systemic risks from any vulnerabilities present. Traditional vulnerability detection methods suffer from low automation and high false positive rates, while existing deep learning-based approaches inadequately extract contract features, thereby limiting detection accuracy. To address these issues, this paper proposes FEMD: a feature-enhanced aided multimodal feature fusion method for smart contract vulnerability detection. Building on the foundation of addressing the low automation of traditional detection tools, our method improves the model's feature extraction performance and detection capabilities. Specifically, we construct a contract graph through Comprehensive Expert-Graph Fusion, combining multi-modal feature fusion using a multi-head attention mechanism with expert patterns to ensure the capture and effective preservation of all critical information during the fusion process. To delve deeper into potential information within smart contract graphs, we employ a Feature Enhancer that leverages transpose operations on feature matrices to extract complex interaction patterns across different dimensions. Our approach is validated through batches of experiments on the Ethereum open dataset focusing on reentrancy and timestamp dependency vulnerabilities, demonstrating significant improvements in detection accuracy and robustness. © 2024 IEEE.

关键词： Deep learning

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：