检索结果-内蒙古大学图书馆

2022 Findings of the Association for Computational Linguistics: EMNLP 2022

作者： Wu, Hongqiu Ding, Ruixue Zhao, Hai Chen, Boli Xie, Pengjun Huang, Fei Zhang, Min Department of Computer Science and Engineering Shanghai Jiao Tong University China Key Laboratory of Shanghai Education Commission for Intelligent Interaction and Cognitive Engineering Shanghai Jiao Tong University China School of Computer Science and Technology Soochow University China Damo Academy Alibaba Group China

Multiple pre-training objectives fill the vacancy of the understanding capability of single-objective language modeling, which serves the ultimate purpose of pre-trained language models (PrLMs), generalizing well on a mass of scenarios. However, learning multiple training objectives in a single model is challenging due to the unknown relative significance as well as the potential contrariety between them. Empirical studies have shown that the current objective sampling in an ad-hoc manual setting makes the learned language representation barely converge to the desired optimum. Thus, we propose MOMETAS, a novel adaptive sampler based on meta-learning, which learns the latent sampling pattern on arbitrary pre-training objectives. Such a design is lightweight with negligible additional training overhead. To validate our approach, we adopt five objectives and conduct continual pre-training with BERT-base and BERT-large models, where MOMETAS demonstrates universal performance gain over other rule-based sampling strategies on 14 natural language processing tasks. © 2022 Association for Computational Linguistics.

关键词： Modeling languages

来源：评论

学校读者我要写书评

暂无评论

LabelEase: A Semi-Automatic Tool for Efficient and Accurate Trace Labeling in Microservices

LabelEase: A Semi-Automatic Tool for Efficient and Accurate ...

引用

International Symposium on Software Reliability Engineering (ISSRE)

作者： Shenglin Zhang Zeyu Che Zhongjie Pan Xiaohui Nie Yongqian Sun Lemeng Pan Dan Pei Nankai University Haihe Laboratory of Information Technology Application Innovation Computer Network Information Center Chinese Academy of Sciences Tianjin Key Laboratory of Software Experience and Human Computer Interaction Huawei Beijing National Research Center for Information Science and Technology Tsinghua University

ISBN: (数字)9798350353884

ISBN: (纸本)9798350353891

Trace data is crucial for system observability and maintainability within microservices architectures, and many operation algorithms depend heavily on trace data, including anomaly detection, root cause analysis, etc. However, the actual performance of these algorithms might be unsatisfactory due to the absence of high-quality labeled datasets for effective training and evaluation. Since billions of traces could be generated daily for large-scale microservices, labeling overhead is the main hurdle to obtaining high-quality trace *** this paper, we propose LabelEase, a novel semi-automatic trace labeling tool, which uses active learning techniques to achieve efficient and accurate trace labeling. For anomaly trace labeling, LabelEase clusters similar traces with a graph-based trace representation technique and selects a few representative traces for human labeling, avoiding labeling most of the traces. For root cause labeling, LabelEase aggregates the labeled anomalous traces and identifies the service’s failures for operators to label. Our systematic experiments on two large-scale datasets show that LabelEase achieves over 0.98 F 1 -score in anomaly trace labeling and 0.89 precision of failure detection in root cause labeling, LabelEase can reduce operators’ labeling overhead by more than 99.9%. To the best of our knowledge, we are the first to propose a semi-automatic trace labeling tool capable of achieving efficient and accurate trace labeling.

关键词： Training Location awareness Accuracy Costs Transfer learning Microservice architectures Manuals Data models Labeling Anomaly detection

来源：评论

学校读者我要写书评

暂无评论

Unveiling Vulnerability of Self-Attention 30

Unveiling Vulnerability of Self-Attention

引用

Joint 30th International Conference on Computational Linguistics and 14th International Conference on Language Resources and Evaluation, LREC-COLING 2024

作者： Liong, Khai Jiet Wu, Hongqiu Zhao, Hai Xiaoshan, Li Minjie, Bian Department of Computer Science and Engineering Shanghai Jiao Tong University China Key Laboratory of Shanghai Education Commission for Intelligent Interaction and Cognitive Engineering Shanghai Jiao Tong University China Shanghai Technology Innovation Department Shanghai Data Group Co. Ltd. China Shanghai Data Group Co. Ltd. China

ISBN: (纸本)9782493814104

Pre-trained language models (PLMs) are shown to be vulnerable to minor word changes, which poses a big threat to real-world systems. While previous studies directly focus on manipulating word inputs, they are limited by their means of generating adversarial samples, lacking generalization to versatile real-world attack. This paper studies the basic structure of transformer-based PLMs, the self-attention (SA) mechanism. (1) We propose a powerful perturbation technique HackAttend, which perturbs the attention scores within the SA matrices via meticulously crafted attention masks. We show that state-of-the-art PLMs fall into heavy vulnerability that minor attention perturbations (1%) can produce a very high attack success rate (98%). Our paper expands the conventional text attack of word perturbations to more general structural perturbations. (2) We introduce S-Attend, a novel smoothing technique that effectively makes SA robust via structural perturbations. We empirically demonstrate that this simple yet effective technique achieves robust performance on par with adversarial training when facing various text attackers. Code is publicly available at ***/liongkj/HackAttend. © 2024 ELRA Language Resource Association: CC BY-NC 4.0.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

How can we can create an equitable CHI

How can we can create an equitable CHI

引用

Extended Abstracts of the 2023 CHI Conference on human Factors in Computing Systems, CHI EA 2023

作者： Grady, Siobahn Day Bauer, Christine Wehbe, Rina Spiel, Katta Muller, Michael Harrington, Christina School of Library and Information Sciences North Carolina Central University DurhamNC United States Department of Information and Computing Sciences Utrecht University Utrecht Netherlands Computer Science Dalhousie University HalifaxNS Canada Faculty of Informatics Tu Wien Vienna Austria Human Centered Ai Ibm Research CambridgeMA United States Human-Computer Interaction Institute Carnegie Mellon University PittsburghPA United States

ISBN: (纸本)9781450394222

This panel aims to generate conversation toward creating a more equitable CHI. In recognizing our community's hard work thus far, this panel seeks to engage panelists and participants with thought-provoking questions to garner and promote actionable items for the community. We intend to have an open dialogue on allyship, diversity, equity, and inclusion to achieve a CHI for all. © 2023 Owner/Author.

关键词： allyship diversity equality equity inclusion

来源：评论

学校读者我要写书评

暂无评论

Forebrain excitatory neuron-specific loss of Brpf1 attenuates excitatory synaptic transmission and impairs spatial and fear memory

引用

Neural Regeneration Research 2024年第5期19卷 1133-1141页

作者： Baicheng Zhao Hang Zhang Ying Liu Gaoyu Zu Yuxiao Zhang Jiayi Hu Shuai Liu Linya You Department of Human Anatomy&Histoembryology School of Basic Medical SciencesFudan UniversityShanghaiChina Shanghai Key Laboratory of Brain Functional Genomics(Ministry of Education) Affiliated Mental Health Center(ECNU)School of Psychology and Cognitive ScienceEast China Normal UniversityShanghaiChina Shanghai Changning Mental Health Center ShanghaiChina NYU-ECNU Institute of Brain and Cognitive Science at NYU Shanghai ShanghaiChina Key Laboratory of Medical Imaging Computing and Computer Assisted Intervention of Shanghai ShanghaiChina

Bromodomain and plant homeodomain(PHD)finger containing protein 1(Brpf1)is an activator and scaffold protein of a multiunit complex that includes other components involving lysine acetyltransferase(KAT)6A/6B/***1,KAT6A,and KAT6B mutations were identified as the causal genes of neurodevelopmental disorders leading to intellectual *** previous work revealed strong and specific expression of Brpf1 in both the postnatal and adult forebrain,especially the hippocampus,which has essential roles in learning and ***,we hypothesized that Brpf1 plays critical roles in the function of forebrain excitatory neurons,and that its deficiency leads to learning and memory *** test this,we knocked out Brpf1 in forebrain excitatory neurons using *** found that Brpf1 deficiency reduced the frequency of miniature excitatory postsynaptic currents and downregulated the expression of genes Pcdhgb1,Slc16a7,Robo3,and Rho,which are related to neural development,synapse function,and memory,thereby damaging spatial and fear memory in *** findings help explain the mechanisms of intellectual impairment in patients with BRPF1 mutation.

关键词： behavioral test Brpf1 CAMKIIa-Cre intellectual disability miniature excitatory postsynaptic current mRNA-Seq

来源：评论

学校读者我要写书评

暂无评论

SIVD: Dataset of Iranian Vehicles for Real-Time Multi-Camera Video Tracking and Recognition

SIVD: Dataset of Iranian Vehicles for Real-Time Multi-Camera...

引用

International Conference of Signal Processing and Intelligent Systems (ICSPIS)

作者： Farbod Siahkali Seyed Amirmahdi Alavi Mehdi Tale Masouleh Human and Robot Interaction Laboratory School of Electrical and Computer Engineering University of Tehran Tehran Iran Human and Robot Interaction Laboratory School of Computer Science Islamic Azad University Central Tehran Branch Tehran Iran

ISBN: (纸本)9781665476249

In this paper, a new publicly available 1 web-Scraped Iranian Vehicle Dataset (SIVD) for simultaneous real-time vehicle tracking and recognition is proposed. The datasets provided for Iranian cars in the literature have two fundamental problems. First, the lack of images from different angles, and second, the small number of classes compared to the dispersion of car models in the real world. Therefore, for the purposes of this paper, Iranian vehicle images from car sales websites are collected, and the SIVD dataset is proposed which contains 29 classes and 36,705 images. This paper aims at developing a classification network for Iranian vehicle recognition and implement a real-time tracking system using the YOLOv5 network to perform real-time vehicle model recognition and tracking tasks simultaneously. The ResNet50 achieved an accuracy of 99.29%, the highest among the investigated classification networks. 1 https://***/Farbod-Siahkali/Iranian-Vehicle-Tracking-and-Recognition

关键词： Representation learning Image recognition Video tracking Machine learning algorithms Neural networks Real-time systems Automobiles

来源：评论

学校读者我要写书评

暂无评论

Stakeholder-Informed Prioritization (SIP): A Technique for Quickly Gauging Research and Design Priorities 25

Stakeholder-Informed Prioritization (SIP): A Technique for Q...

引用

Proceedings of the 24th interaction Design and Children

作者： Qiao Jin Leah Ajmani Samantha Singh Adhitya Balasubramanian Svetlana Yarosh Human-Computer Interaction Institute Carnegie Mellon University Pittsburgh Pennsylvania USA Department of Computer Science and Engineering University of Minnesota Minneapolis Minnesota USA The Blake School Minneapolis Minnesota USA Eagan High School Eagan Minnesota USA

来源：评论

学校读者我要写书评

暂无评论

A Dual-view Semi-supervised Learning Framework for Combinatorial Few-shot Fault Diagnosis

A Dual-view Semi-supervised Learning Framework for Combinato...

引用

IEEE International Conference on Performance, Computing and Communications (IPCCC)

作者： Yumeng Liu Wenzhang Zhong Ke Yan Ling Tian Beijing Key Laboratory of Human-Computer Interaction Institute of Software Chinese Academy of Sciences Beijing China School of Computer Science and Technology University of Chinese Academy of Sciences Beijing China School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu China

The accurate and comprehensive fault diagnosis is pivotal for the maintenance of mechanical equipment, which may directly impact the safety in production and avoid unnecessary losses of equipment. With the pervasively adopted sensing components, sensing data can be collected from mechanical equipment and analyzed for fault diagnosis. Previous solutions apply sophisticated deep learning models for such fault diagnosis tasks and have achieved impressive performance. However, these methods usually ignore the existences of diverse characteristics of faults, and formulate the fault diagnosis as a simple classification problem. Considering necessity of comprehensive and multi-view fault diagnosis, this work proposes a novel dual-view learning framework for combinatorial mechanical fault classification. The framework includes dual encoders to simultaneously learn the feature representation tied with the position and size of faults on equipment. Then these features are mutually mixed for better representation and the mixed features are merged and feed into deep neural networks for further representation learning. The query samples are classified through metric-based comparison to search for the closest type of composite fault. Moreover, the proposed framework also applies typical metric-based meta learning method to handle the issue of small training datasets, and adopts semi-supervised learning method to make use of both labeled and unlabeled samples. Finally, the experimental results on public datasets show that the proposed method can outperform previous solutions and achieve state-of-the-art performance.

关键词：

来源：评论

学校读者我要写书评

暂无评论

AI Literacy: Finding Common Threads between Education, Design, Policy, and Explainability

AI Literacy: Finding Common Threads between Education, Desig...

引用

Extended Abstracts of the 2023 CHI Conference on human Factors in Computing Systems, CHI EA 2023

作者： Long, Duri Roberts, Jessica Magerko, Brian Holstein, Kenneth Dipaola, Daniella Martin, Fred Department of Communication Studies Northwestern University United States School of Interactive Computing Georgia Tech United States School of Literature Media and Communication Georgia Tech United States Human-Computer Interaction Institute Carnegie Mellon University United States Personal Robots Group Mit Media Lab United States Computer Science University of Massachusetts Lowell United States

ISBN: (纸本)9781450394222

Fostering public AI literacy has been a growing area of interest at CHI for several years, and a substantial community is forming around issues such as teaching children how to build and program AI systems, designing learning experiences to broaden public understanding of AI, developing explainable AI systems, understanding how novices make sense of AI, and exploring the relationship between public policy, ethics, and AI literacy. Previous workshops related to AI literacy have been held at other conferences (e.g., SIGCSE, AAAI) that have been mostly focused on bringing together researchers and educators interested in AI education in K-12 classroom environments, an important subfield of this area. Our workshop seeks to cast a wider net that encompasses both HCI research related to introducing AI in K-12 education and also HCI research that is concerned with issues of AI literacy more broadly, including adult education, interactions with AI in the workplace, understanding how users make sense of and learn about AI systems, research on developing explainable AI (XAI) for non-expert users, and public policy issues related to AI literacy. © 2023 Owner/Author.

关键词： Public policy

来源：评论

学校读者我要写书评

暂无评论

A Neural Architecture Search Method with Multi-Dimensional Correlation Representation for Fault Diagnosis

A Neural Architecture Search Method with Multi-Dimensional C...

引用

Smart World Congress (SWC), IEEE

作者： Yumeng Liu Tingqi Wang Xu Zheng Ling Tian Hongan Wang Beijing Key Laboratory of Human-Computer Interaction Institute of Software Chinese Academy of Sciences Beijing China School of Computer Science and Technology University of Chinese Academy of Sciences Beijing China School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu China

As deep learning has become more widely used for fault diagnosis, the shortcomings of model transferability and human model design costs are growing increasingly evident. The current work has tackled each of these two issues. Domain adaptation methods can provide transportability by mapping data to a common domain. For the consumption of manually constructed networks, Neural Architecture Search (NAS) is an intelligent solution that automatically select the ideal network structure. Unfortunately, there is little work to connect these two methodologies in fault diagnosis. Furthermore, these approaches are incapable in extracting internal correlations from instrumental signal data. In this paper, we propose a transferrable NAS fault diagnosis method, which maximizes savings in manual costs. We also integrate graph neural networks (GNN) in a stepwise manner into the NAS and domain adaptation to provide a multi-dimension correlation representation. Firstly, we structure the signal data into graph based on temporal relationships. Then, NAS designs automatically for selecting the optimal network structure and useful features. In NAS, graph encoder of GNN is used to establish internal relationships. Finally, domain adaptation provides transferability. At the same time, GNN’s weights in domain adaptation help to further mine data internal associations. We tested our method on two publicly available datasets and performed comparisons. The experimental results show the proposed method has superior performance.

关键词：

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：