检索结果-内蒙古大学图书馆

48th IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2023

作者： Zhang, Shun Li, Tongliang Bai, Jiaqi Li, Zhoujun Beihang University State Key Lab of Software Development Environment Beijing China

ISBN: (纸本)9781728163277

Out-of-Domain (OOD) detection from user utterances plays an important part in task-oriented dialogue systems. Recent studies utilize supervised or self-supervised contrastive learning (CL) to learn discriminative semantic features for OOD detection. The supervised contrastive learning (SCL) methods only model class-level features of different in-domain (IND) intents while self-supervised CL (SSCL) methods can model instance-level features. However, the SSCL methods require complex data augmentation and are vulnerable to intrinsic false-negative pairs. To address the issues above and leverage both types of CL, we propose a novel Label-Guided Contrastive Learning (LGCL) framework. LGCL models both instance-level and class-level discriminative semantic representations by employing the sample and its corresponding label simultaneously, such that the prior knowledge of IND can be fully leveraged. Experiment and analysis results on two public benchmark datasets show that the proposed method significantly outperforms baselines on the OOD detection task. © 2023 IEEE.

关键词： Contrastive Learning Intent Detection Out-of-Domain

来源：评论

学校读者我要写书评

暂无评论

Adaptive Token Selection and Fusion Network for Multimodal Sentiment Analysis 1

引用

30th International Conference on MultiMedia Modeling, MMM 2024

作者： Li, Xiang Lu, Ming Guo, Ziming Zhang, Xiaoming School of Cyber Science and Technology Beihang University Beijing China State Key Laboratory of Software Development Environment Beihang University Beijing100191 China

ISBN: (数字)9783031533112

ISBN: (纸本)9783031533105

Multimodal sentiment analysis aims to predict human sentiment polarity with multiple modalities. Most existing methods focus on directly integrating original modal features into multimodal fusion, ignoring the redundancy and heterogeneity across modalities. In this paper, we propose a simple but efficient Adaptive Token Selection and Fusion Network (ATSFN) to mitigate the effect of redundancy and heterogeneity. ATSFN employs adaptive trainable tokens to extract unimodal informative tokens and perform dynamic multimodal token fusion. Specifically, we first integrate critical information from original features into adaptive selection tokens through token selection transformers. Sentiment features flow through these smaller sequences of tokens to capture important information while reducing redundancy. Next, we introduce a token fusion transformer to fuse multimodal features dynamically. It adaptively estimates the unique contribution of each modality to sentiment tendencies through learnable fusion tokens. Experiments on two benchmark datasets demonstrate that our proposed approach achieves competitive performance and significant improvements. © 2024, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Sentiment analysis

来源：评论

学校读者我要写书评

暂无评论

Towards optimized tensor code generation for deep learning on sunway many-core processor

引用

Frontiers of Computer Science 2024年第2期18卷 1-15页

作者： Mingzhen LI Changxi LIU Jianjin LIAO Xuegui ZHENG Hailong YANG Rujun SUN Jun XU Lin GAN Guangwen YANG Zhongzhi LUAN Depei QIAN State Key Laboratory of Software Development Environment Beijing 100191China School of Computer Science and Engineering Beihang UniversityBeijing 100191China National University of Singapore Singapore 119077Singapore State Key Laboratory of Mathematical Engineering and Advanced Computing Wuxi 214000China Science and Technology on Special System Simulation Laboratory Beijing Simulation Center Beijing 100854China Department of Computer Science and Technology Tsinghua UniversityBeijing 100084China

The flourish of deep learning frameworks and hardware platforms has been demanding an efficient compiler that can shield the diversity in both software and hardware in order to provide application *** the existing deep learning compilers,TVM is well known for its efficiency in code generation and optimization across diverse hardware *** the meanwhile,the Sunway many-core processor renders itself as a competitive candidate for its attractive computational power in both scientific computing and deep learning *** paper combines the trends in these two ***,we propose swTVM that extends the original TVM to support ahead-of-time compilation for architecture requiring cross-compilation such as *** addition,we leverage the architecture features during the compilation such as core group for massive parallelism,DMA for high bandwidth memory transfer and local device memory for data locality,in order to generate efficient codes for deep learning workloads on *** experiment results show that the codes generated by swTVM achieve 1.79x improvement of inference latency on average compared to the state-of-the-art deep learning framework on Sunway,across eight representative *** work is the first attempt from the compiler perspective to bridge the gap of deep learning and Sunway processor particularly with productivity and efficiency in *** believe this work will encourage more people to embrace the power of deep learning and Sunwaymany-coreprocessor.

关键词： sunway processor deep learning compiler code generation performance optimization

来源：评论

学校读者我要写书评

暂无评论

Patterns and mechanism of wintertime penetrating fronts in the East China Sea

引用

Science China Earth Sciences 2024年第11期67卷 3500-3514页

作者： Peng YE Daji HUANG Jiliang XUAN Shuangyan HE Ocean College Zhejiang UniversityZhoushan 316021China State Key Laboratory of Satellite Ocean Environment Dynamics Second Institute of OceanographyMinistry of Natural ResourcesHangzhou 310012China Guangxi Key Laboratory of Beibu Gulf Marine Resources Environment and Sustainable DevelopmentFourth Institute of OceanographyMinistry of Natural ResourcesBeihai 536015China

Penetrating fronts play an important role in the cross-shelf transport of terrestrial materials in the East China Sea(ECS). Using long-term satellite remote sensing data and numerical simulation data, the most likely period of occurrence and region of the penetrating fronts in the western ECS are analyzed in this study, and the evolutionary process and mechanism are also investigated. The statistical results of satellite-derived chlorophyll data from 1998 to 2022(25 years) reveal that penetrating fronts occur most frequently near 27°N in winter, with the frequencies of occurrence in January, February and March being 47%,65% and 64%, respectively. Backward Lagrangian tracer experiments demonstrate that the penetrating water near 27°N originate from three different regions. The core penetrating low-salinity water originates from the Zhejiang-Fujian coast, while the northern marginal water originates from the northern part of Taiwan Island, and the southern marginal water originates from the central Taiwan Strait. Mechanism analysis reveals that the core penetrating low-salinity water evolves in three successive ***, under the influence of the Zhejiang-Fujian Coastal Current driven by strong northeasterly winds, the low-salinity water along the Zhejiang-Fujian coast moves southwestward parallel to the coastline and accumulates in the western Taiwan ***, during the relaxation stage of the northeasterly wind, the northeastward Taiwan Strait Current strengthens, and the lowsalinity water accumulated in the western Taiwan Strait moves to the northeast. Third, when the northeastward-moving lowsalinity water runs into the Western Kuroshio Branch off the northeastern Taiwan Island, the low-salinity water is rapidly stretched eastward, and a significant penetrating front eventually appears. Since the synoptic northeasterly wind and the Western Kuroshio Branch are dominant dynamic factors in this region during wintertime, we believe that the synoptic winterti

关键词： Penetrating fronts Satellite-derived chlorophyll Lagrangian tracer experiments Synoptic-scale fluctuation Western Kuroshio Branch

来源：评论

学校读者我要写书评

暂无评论

QAAS:quick accurate auto-scaling for streaming processing

引用

Frontiers of Computer Science 2024年第1期18卷 5-15页

作者： Shiyuan LIU Yunchun LI Hailong YANG Ming DUN Chen CHEN Huaitao ZHANG Wei LI State Key Laboratory of Software Development Environment Beijing 100191China School of Computer Science and Engineering Beihang UniversityBeijing 100191China School of Cyber Science and Technology Beihang UniversityBeijing 100191China School of Computer Science and Technology Xi’an Jiaotong UniversityXi’an 710049China

In recent years,the demand for real-time data processing has been increasing,and various stream processing systems have *** the amount of data input to the stream processing system fluctuates,the computing resources required by the stream processing job will also *** resources used by stream processing jobs need to be adjusted according to load changes,avoiding the waste of computing *** present,existing works adjust stream processing jobs based on the assumption that there is a linear relationship between the operator parallelism and operator resource consumption(e.g.,throughput),which makes a significant deviation when the operator parallelism *** paper proposes a nonlinear model to represent operator *** divide the operator performance into three stages,the Non-competition stage,the Non-full competition stage,and the Full competition *** our proposed performance model,given the parallelism of the operator,we can accurately predict the CPU utilization and operator *** with actual experiments,the prediction error of our model is below 5%.We also propose a quick accurate auto-scaling(QAAS)method that uses the operator performance model to implement the auto-scaling of the operator parallelism of the Flink *** to previous work,QAAS is able to maintain stable job performance under load changes,minimizing the number of job adjustments and reducing data backlogs by 50%.

关键词： stream processing performance model auto-scaling

来源：评论

学校读者我要写书评

暂无评论

Minimizing the cost of periodically replicated systems via model and quantitative analysis

引用

Frontiers of Computer Science 2024年第5期18卷 11-24页

作者： Chenhao ZHANG Liang WANG Limin XIAO Shixuan JIANG Meng HAN Jinquan WANG Bing WEI Guangjun QIN State Key Laboratory of Software Development Environment Beihang UniversityBeijing 100191China School of Computer Science and Engineering Beihang UniversityBeijing 100191China School of Cyberspace Security Hainan UniversityHaikou 570228China Smart City College Beijing Union UniversityBeijing 100101China

Geographically replicating objects across multiple data centers improves the performance and reliability of cloud storage *** consistent replicas comes with high synchronization costs,as it faces more expensive WAN transport prices and increased *** replication is the widely used technique to reduce the synchronization *** replication strategies in existing cloud storage systems are too static to handle traffic changes,which indicates that they are inflexible in the face of unforeseen loads,resulting in additional synchronization *** propose quantitative analysis models to quantify consistency and synchronization cost for periodically replicated systems,and derive the optimal synchronization period to achieve the best tradeoff between consistency and synchronization *** on this,we propose a dynamic periodic synchronization method,Sync-Opt,which allows systems to set the optimal synchronization period according to the variable load in clouds to minimize the synchronization *** results demonstrate the effectiveness of our *** with the policies widely used in modern cloud storage systems,the Sync-Opt strategy significantly reduces the synchronization cost.

关键词： periodic replication consistency maintenance synchronization cost synchronization strategy

来源：评论

学校读者我要写书评

暂无评论

Propagation criterion of hydraulic fracture in rock based on the rock micro-cracking mechanism

引用

International Journal of Mining Science and Technology 2025年第3期35卷 433-449页

作者： Qingwang Cai Bingxiang Huang Xinglong Zhao Yuekun Xing School of Resources Environment and Safety Engineering University of South ChinaHengyang 421001 China State Key Laboratory for Fine Exploration and Intelligent Development of Coal Resources China University of Mining and TechnologyXuzhou 221116 China

Hydraulic fracture(HF)formed in rock significantly helps with the development of geo-energy and *** HF formation condition was challenging to understand,with obscure rock micro-cracking mechanisms being a key *** rock micro-cracking mechanism under gradient pore water pressure was analyzed on the scale of mineral particles and it was combined with macroscopic boundary conditions of rock hydraulic fracturing,obtaining the propagation criterion of HF in rock based on the rock micro-cracking mechanism which was verified by *** results show that the disturbed skeleton stress induced by the disturbance of gradient pore water pressure in rock equals the pore water pressure *** overall range of the defined mechanical shape factor a/b is around 1,but greater *** the combined influence of pore water pressure differences and macroscopic boundary stresses on the rock micro-cracking,micro-cracks form among rock mineral particles,micro-cracks connect to form micro-hydraulic fracture surfaces,and micro-hydraulic fracture surfaces open to form macrohydraulic *** begins to form at the micro-cracking initiation pressure(MCIP),which was tested by keeping the HF tip near the initiation *** theoretical value of MCIP calculated by the proposed propagation criterion is close to MCIP tested.

关键词： Hydraulic fracture Propagation criterion Micro-cracking mechanism Pore pressure Stress singularity

来源：评论

学校读者我要写书评

暂无评论

A method for detecting abnormal users with fake stars 28

A method for detecting abnormal users with fake stars

引用

28th International DMS Conference on Visualization and Visual Languages, DMSVIVA 2022

作者： Jiang, Jing Li, Hao Liu, Yifan Zhang, Li The State Key Laboratory of Software Development Environment Beihang University Beijing China

ISBN: (纸本)1891706551

In GitHub, users star interesting repositories, and the number of stars is viewed as the significant measure of repository popularity. Some repositories obtain fake stars by unjustified means, which ruin efforts that communities have made stars a valuable indicator, and bring negative impacts in GitHub. Therefore, it is important to stop abusing GitHub stars and detect abnormal users who provide fake stars. In this paper, we first define features from the user dimension and repository dimension. Then we perform differential analysis and find that most of the features show a significant difference between abnormal users and normal users. Next, we propose a method AUDetec for Abnormal User Detection. The method AUDetec uses the decision tree to detect the abnormal users based on two features, including the sum of repositories starred by the user and the median value of the number of days since creation for repositories starred by the user. We evaluate the effectiveness of AUDetec on the data set which contains 120 abnormal users and 240 normal users. The experiment results show that AUDetec has a high performance by achieving an accuracy of 99.86% on average. © 2022 by KSI Research Inc. and Knowledge Systems Institute, USA.

关键词： Open source software

来源：评论

学校读者我要写书评

暂无评论

Hyperbolic distance guided multi-positive graph contrastive learning 23

Hyperbolic distance guided multi-positive graph contrastive ...

引用

15th International Conference on Machine Learning and Computing, ICMLC 2023

作者： Liu, Yanxi Lang, Bo State Key Laboratory of Software Development Environment Beihang University Beijing Shi Haidian Qu China Zhongguancun Laboratory Beijing Shi China

ISBN: (纸本)9781450398411

Graph contrastive learning (GCL) aims to obtain embeddings for graph data in an unsupervised manner. Most existing GCL models are designed for homogeneous graphs, whereas the two critical aspects of heterogeneous GCL models are data augmentation and contrastive sample selection. In this paper, we propose a contrastive learning model for heterogeneous graphs. We take advantage of the property that hyperbolic space is suitable for representing highly hierarchical graph data. We use the normalized hyperbolic distance as an indicator to adaptively select edges to remove in data augmentation to obtain different graph views. In addition, we propose a positive example expansion strategy that integrates topological information and attributes information to obtain more highly relevant positive samples and provide more contrasting information. We conduct node classification experiments on several public heterogeneous graph datasets to evaluate the performance of our model. The results show that our model outperforms all unsupervised comparison methods and even outperforms the best-supervised model on several datasets to reach state-of-the-art. © 2023 ACM.

关键词： Graph neural networks

来源：评论

学校读者我要写书评

暂无评论

MissionML: A Mission Architecture Modeling Language based on Unified Architecture Framework

引用

INCOSE International Symposium 2024年第1期34卷 479-502页

作者： Yang, Yilong Song, Wenxiao Chang, Chuangye Gao, Xinghai State Key Laboratory of Complex & Critical Software Environment School of Software Beihang University Beijing China School of Software Beihang University Beijing China Institute of Unmanned Systems Beihang University Beijing China

The missions of complex systems, organizations, or groups can be identified through careful requirements and domain analysis. Mission architecture modeling is a crucial step for enterprise modeling and design. However, mission modeling is lacking from the Unified Architecture Framework (UAF), which system engineers must specify and model from sketch. This paper proposes a Mission Architecture Modeling Language (MissionML). This two-layer architecture language generalizes the general common and special knowledge from four typical missions as a core layer and mission layer. The elements of the core layer are obtained through slight extensions of the operational layer elements in UAFML, and the elements of the mission layer are derived from the elements in the core layer. Moreover, MissionML is implemented as a UAF profile, incorporating numerous domain concepts in its syntax and semantics for mission modeling. Finally, we use five public case studies to demonstrate learnability and extensibility from the viewpoint of system engineers. Copyright © 2024 by Yilong Yang, Wenxiao Song, Chuangye Chang, and Xinghai Gao. Permission granted to INCOSE to publish and use.

关键词： Unified Modeling Language

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：