检索结果-内蒙古大学图书馆

36th Conference on Neural Information Processing Systems, NeurIPS 2022

作者： Luo, Shengjie Li, Shanda Zheng, Shuxin Liu, Tie-Yan Wang, Liwei He, Di National Key Laboratory of General Artificial Intelligence School of Intelligence Science and Technology Peking University China Machine Learning Department School of Computer Science Carnegie Mellon University United States Microsoft Research United States Center for Data Science Peking University China Zhejiang Lab China

ISBN: (纸本)9781713871088

Relative Positional Encoding (RPE), which encodes the relative distance between any pair of tokens, is one of the most successful modifications to the original Transformer. As far as we know, theoretical understanding of the RPE-based Transformers is largely unexplored. In this work, we mathematically analyze the power of RPE-based Transformers regarding whether the model is capable of approximating any continuous sequence-to-sequence functions. One may naturally assume the answer is in the affirmative-RPE-based Transformers are universal function approximators. However, we present a negative result by showing there exist continuous sequence-to-sequence functions that RPE-based Transformers cannot approximate no matter how deep and wide the neural network is. One key reason lies in that most RPEs are placed in the softmax attention that always generates a right stochastic matrix. This restricts the network from capturing positional information in the RPEs and limits its capacity. To overcome the problem and make the model more powerful, we first present sufficient conditions for RPE-based Transformers to achieve universal function approximation. With the theoretical guidance, we develop a novel attention module, called Universal RPE-based (URPE) Attention, which satisfies the conditions. Therefore, the corresponding URPE-based Transformers become universal function approximators. Extensive experiments covering typical architectures and tasks demonstrate that our model is parameter-efficient and can achieve superior performance to strong baselines in a wide range of applications. The code will be made publicly available at https://***/lsj2408/URPE. © 2022 Neural information processing systems foundation. All rights reserved.

关键词： Encoding (symbols)

来源：评论

学校读者我要写书评

暂无评论

Identity-Driven Hierarchical Role-Playing Agents

arXiv

引用

arXiv 2024年

作者： Sun, Libo Wang, Siyuan Huang, Xuanjing Wei, Zhongyu School of Data Science Fudan University China University of Southern California United States School of Computer Science Fudan University China Research Institute of Intelligent Complex Systems Fudan University China

Utilizing large language models (LLMs) to achieve role-playing has gained great attention recently. The primary implementation methods include leveraging refined prompts and fine-tuning on role-specific datasets. However, these methods suffer from insufficient precision and limited flexibility respectively. To achieve a balance between flexibility and precision, we construct a Hierarchical Identity Role-Playing Framework (HIRPF) based on identity theory, constructing complex characters using multiple identity combinations. We develop an identity dialogue dataset for this framework and propose an evaluation benchmark including scale evaluation and open situation evaluation. Empirical results indicate the remarkable efficacy of our framework in modeling identity-level role simulation, and reveal its potential for application in social simulation. The source code is available in https://***/r/HIS-7DBB. Copyright © 2024, The Authors. All rights reserved.

关键词： Benchmarking

来源：评论

学校读者我要写书评

暂无评论

Enhancing action discrimination via category-specific frame clustering for weakly-supervised temporal action localization

引用

Frontiers of Information Technology & Electronic Engineering 2024年第6期25卷 809-823页

作者： Huifen XIA Yongzhao ZHAN Honglin LIU Xiaopeng REN School of Computer Science and Communication Engineering Jiangsu UniversityZhenjiang 212013China Jiangsu Engineering Research Center of Big Data Ubiquitous Perception and Intelligent Agricultural Applications Zhenjiang 212013China Changzhou Vocational Institute of Mechatronic Technology Changzhou 213164China

Temporal action localization (TAL) is a task of detecting the start and end timestamps of action instances and classifying them in an untrimmed video. As the number of action categories per video increases, existing weakly-supervised TAL (W-TAL) methods with only video-level labels cannot provide sufficient supervision. Single-frame supervision has attracted the interest of researchers. Existing paradigms model single-frame annotations from the perspective of video snippet sequences, neglect action discrimination of annotated frames, and do not pay sufficient attention to their correlations in the same category. Considering a category, the annotated frames exhibit distinctive appearance characteristics or clear action ***, a novel method to enhance action discrimination via category-specific frame clustering for W-TAL is proposed. Specifically,the K-means clustering algorithm is employed to aggregate the annotated discriminative frames of the same category, which are regarded as exemplars to exhibit the characteristics of the action category. Then, the class activation scores are obtained by calculating the similarities between a frame and exemplars of various categories. Category-specific representation modeling can provide complimentary guidance to snippet sequence modeling in the mainline. As a result, a convex combination fusion mechanism is presented for annotated frames and snippet sequences to enhance the consistency properties of action discrimination,which can generate a robust class activation sequence for precise action classification and localization. Due to the supplementary guidance of action discriminative enhancement for video snippet sequences, our method outperforms existing single-frame annotation based methods. Experiments conducted on three datasets (THUMOS14, GTEA, and BEOID) show that our method achieves high localization performance compared with state-of-the-art methods.

关键词： Weakly supervised Temporal action localization Single-frame annotation Category-specific Action discrimination

来源：评论

学校读者我要写书评

暂无评论

Enhancing Cloud IoT Security With Blockchain and SDN

Enhancing Cloud IoT Security With Blockchain and SDN

引用

Communication, computer sciences and Engineering (IC3SE), International Conference on

作者： Kumaran U Thammisetty Swetha Ishi Sharma R Keerthana V Tejashwini Ganesh Gopal Devarajan Department of Computer Science & Engineering Amrita School of Computing Amrita Vishwa Vidyapeetham Bengaluru India Department of Computer Science & Engineering (Data Science) Madanapalle Institute of Technology & Science Madanapalle Andhra Pradesh India Department of Electronics and Communication Engineering Amrita School of Engineering Amrita Vishwa Vidyapeetham Bengaluru India Department of Computer Science and Engineering SRM Institute of Science and Technology Delhi NCR campus Ghaziabad Uttar Pradesh India

ISBN: (数字)9798350366846

ISBN: (纸本)9798350366853

In the fast changing world of cloud based internet of things (IoT) eco-systems, secure connectivity between devices and cloud services is vital. On the other hand, the rising Man-in-the- Middle (MitM) attacks present serious threats to data integrity and confidentiality. In this regard, through a combination of Software Defined Networking (SDN) and Blockchain technology, this study proposes an innovative technique for improving security architecture in cloud IoT settings against MitM attacks. The solution put forward relies on blockchain to provide immutable transactions that are used as tamper-proof ledgers for device authentication and secure data sharing. Smart contracts create unchangeable transactions that also maintain the anonymity of devices and their communication processes. Additionally, SDN technology segments traffic, implements dynamic network policies, and therefore reduces MitM attack vectors through traffic inspection and access control. The testing process has shown that the approach effectively prevents MitM attacks while ensuring safe communication channels in cloud-based IoT systems. This implementation demonstrates improved data integrity, confidentiality, and security against unwanted intrusion attempts. This research paper contributes to the development of the cloud-based IoT security paradigm and highlights the synergistic potential of Blockchain and SDN in dealing with common security issues and improving the trustworthiness of networked devices for the Internet of Things.

关键词： Cloud computing data integrity Smart contracts computer architecture Inspection Vectors Blockchains

来源：评论

学校读者我要写书评

暂无评论

Simulation of Photonic Metasurfaces using a Surface Integral Equation Solver with Generalized Sheet Transition Conditions

Simulation of Photonic Metasurfaces using a Surface Integral...

引用

Antennas and Propagation (LACAP), IEEE Latin American Conference on

作者： Sebastian Celis Sierra Partha Mondal Ran Zhao Hakan Bagci Electrical and Computer Engineering (ECE) Program King Abdullah University of Science and Technology Thuwal Saudi Arabia School of Electronic Science and Engineering University of Electronic Science and Technology of China Chengdu China

ISBN: (数字)9798331527587

ISBN: (纸本)9798331527594

Photonic metasurfaces are efficiently and accurately simulated using a surface integral equation (SIE) solver. This solver models the metasurface as an infinitesimally thin sheet on which generalized sheet transition conditions (GSTCs) are enforced. GSTCs connect the electromagnetic fields on both sides of the metasurface using equivalent susceptibility tensors and avoid modelling of geometrically intricate metasurface unit cells by the SIE solver. Susceptibility tensors are obtained from the transmission and the reflection coefficients associated with a unit cell. A numerical example, involving a metasurface that supports high-quality factor resonances and excited by a Gaussian beam, is presented to demonstrate the accuracy of the proposed approach.

关键词： Tensors Accuracy Integral equations Metasurfaces Mathematical models Numerical models Reflection coefficient Electromagnetic fields Dispersion Photonics

来源：评论

学校读者我要写书评

暂无评论

MathEagle: Accurate prediction of drug-drug interaction events via multi-head attention and heterogeneous attribute graph learning

引用

computers in Biology and Medicine 2024年 177卷 108642-108642页

作者： Hou, Lin-Xuan Yi, Hai-Cheng You, Zhu-Hong Chen, Shi-Hong Zheng, Jia Kwoh, Chee Keong School of Computer Science Northwestern Polytechnical University Xi'an710129 China Key Laboratory of Big Data Storage and Management Northwestern Polytechnical University Ministry of Industry and Information Technology Xi'an710129 China School of Computer Science and Engineering Nanyang Technological University 50 Nanyang Avenue Singapore

Background: Drug-drug interaction events influence the effectiveness of drug combinations and can lead to unexpected side effects or exacerbate underlying diseases, jeopardizing patient prognosis. Most existing methods are restricted to predicting whether two drugs interact or the type of drug-drug interactions, while very few studies endeavor to predict the specific risk levels of side effects of drug combinations. Methods: In this study, we propose MathEagle, a novel approach to predict accurate risk levels of drug combinations based on multi-head attention and heterogeneous attribute graph learning. Initially, we model drugs and three distinct risk levels between drugs as a heterogeneous information graph. Subsequently, behavioral and chemical structure features of drugs are utilized by message passing neural networks and graph embedding algorithms, respectively. Ultimately, MathEagle employs heterogeneous graph convolution and multi-head attention mechanisms to learn efficient latent representations of drug nodes and estimates the risk levels of pairwise drugs in an end-to-end manner. Results: To assess the effectiveness and robustness of the model, five-fold cross-validation, ablation experiments, and case studies were conducted. MathEagle achieved an accuracy of 85.85 % and an AUC of 0.9701 on the drug risk level prediction task and is superior to all comparative models. The MathEagle predictor is freely accessible at http://120.77.11.78/MathEagle/. Conclusions: The experimental results indicate that MathEagle can function as an effective tool for predicting accurate risk of drug combinations, aiding in guiding clinical medication, and enhancing patient outcomes. © 2024 Elsevier Ltd

关键词： Message passing

来源：评论

学校读者我要写书评

暂无评论

An Algorithm for Extracting Image Features Using a Random Deep Neural Network 3

An Algorithm for Extracting Image Features Using a Random De...

引用

3rd International Conference on Advance Computing and Innovative Technologies in Engineering, ICACITE 2023

作者： Dharmadhikari, Sonali Prasad Rabi, J. Karetla, Girija Rani Joshi, Kapil Shivakumar, K. Tiwari, Mohit Institute of Management and Entrepreneurship Development Pune India Mangalam College of Engineering Department of Mechanical Engineering Kerala Kottayam India We Stern Sydney University School of Computer Data and Mathematical Sciences Australia Uttaranchal Institute of Technology Uttaranchal University India SR University School of Computer Science and Artificial Intelligence Telangana Warangal India Bharati Vidyapeeth's College of Engineering Department of Computer Science and Engineering A-4 Rohtak Road Psachim Vihar Delhi India

ISBN: (纸本)9798350399264

In this paper, a random deep convolutional neural network-based technique for extracting image features is devised and put into practice. The image features are also richer following the image preprocessing procedure described in this article. Additionally, more consistent feature points are recognized, duplicate local features are deleted, and this prevents the feature matching from encountering too many unmatched pairs. Since the self-encoder can successfully rebuild the training data set, data that is close to the training set, and the data itself, this work offers some original ideas for constructing the model. The study's findings have demonstrated the successful outcomes. © 2023 IEEE.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

DivLog: Log Parsing with Prompt Enhanced In-Context Learning

DivLog: Log Parsing with Prompt Enhanced In-Context Learning

引用

International Conference on Software Engineering (ICSE)

作者： Junjielong Xu Ruichun Yang Yintong Huo Chengyu Zhang Pinjia He School of Data Science The Chinese University of Hong Kong Shenzhen China Dept. of Computer Science & Engineering The Chinese University of Hong Kong China ETH Zurich Switzerland Shenzhen Research Institute of Big Data China

ISBN: (数字)9798400702174

ISBN: (纸本)9798350382143

Log parsing, which involves log template extraction from semi-structured logs to produce structured logs, is the first and the most critical step in automated log analysis. However, current log parsers suffer from limited effectiveness for two reasons. First, traditional data-driven log parsers solely rely on heuristics or handcrafted features designed by domain experts, which may not consistently perform well on logs from diverse systems. Second, existing super-vised log parsers require model tuning, which is often limited to fixed training samples and causes sub-optimal performance across the entire log source. To address this limitation, we propose Di-vLog, an effective log parsing framework based on the in-context learning (ICL) ability of large language models (LLMs). Specifically, before log parsing, DivLog samples a small amount of offline logs as candidates by maximizing their diversity. Then, during log parsing, DivLog selects five appropriate labeled candidates as examples for each target log and constructs them into a prompt. By mining the semantics of examples in the prompt, DivLog generates a target log template in a training-free manner. In addition, we design a straightforward yet effective prompt format to extract the output and enhance the quality of the generated log templates. We conducted experiments on 16 widely-used public datasets. The results show that DivLog achieves (1) 98.1% Parsing Accuracy, (2) 92.1% Precision Template Accuracy, and (3) 92.9% Recall Template Accuracy on average, exhibiting state-of-the-art performance.

关键词： Training Measurement Analytical models Semantics Feature extraction Tuning Software engineering

来源：评论

学校读者我要写书评

暂无评论

Note on a New Construction of Kantorovich Form q-Bernstein Operators Related to Shape Parameter λ

引用

computer Modeling in Engineering & sciences 2022年第3期130卷 1479-1493页

作者： Qingbo Cai Resat Aslan Fujian Provincial Key Laboratory of Data-Intensive Computing Key Laboratory of Intelligent Computing and Information ProcessingSchool of Mathematics and Computer ScienceQuanzhou Normal UniversityQuanzhou362000China Department of Mathematics Faculty of Sciences and ArtsHarran UniversitySanlıurfa63300Turkey

The main purpose of this paper is to introduce some approximation properties of a Kantorovich kind q-Bernstein operators related to B′ezier basis functions with shape parameterλ∈[−1,1].Firstly,we compute some basic results such as moments and central moments,and derive the Korovkin type approximation theorem for these ***,we estimate the order of convergence in terms of the usual modulus of continuity,for the functions belong to Lipschitz-type class and Peetre’s K-functional,***,with the aid of Maple software,we present the comparison of the convergence of these newly defined operators to the certain function with some graphical illustrations and error estimation table.

关键词： q-calculus (λ,q)-Bernstein polynomials order of convergence Lipschitz-type function Peetre’s K-functional

来源：评论

学校读者我要写书评

暂无评论

Train Faster, Perform Better: Modular Adaptive Training in Over-Parameterized Models

arXiv

引用

arXiv 2024年

作者： Shi, Yubin Chen, Yixuan Dong, Mingzhi Yang, Xiaochen Li, Dongsheng Wang, Yujiang Dick, Robert Lv, Qin Zhao, Yingying Yang, Fan Lu, Tun Gu, Ning Shang, Li China and Shanghai Key Laboratory of Data Science School of Computer Science Fudan University China School of Mathematics Statistics University of Glasgow United Kingdom Microsoft Research Asia Shanghai China Department of Engineering Science University of Oxford United Kingdom Department of Electrical Engineering and Computer Science University of Michigan United States Department of Computer Science University of Colorado Boulder United States School of Microelectronics Fudan University China

Despite their prevalence in deep-learning communities, over-parameterized models convey high demands of computational costs for proper training. This work studies the fine-grained, modular-level learning dynamics of over-parameterized models to attain a more efficient and fruitful training strategy. Empirical evidence reveals that when scaling down into network modules, such as heads in self-attention models, we can observe varying learning patterns implicitly associated with each module's trainability. To describe such modular-level learning capabilities, we introduce a novel concept dubbed modular neural tangent kernel (mNTK), and we demonstrate that the quality of a module's learning is tightly associated with its mNTK's principal eigenvalue λmax. A large λmax indicates that the module learns features with better convergence, while those miniature ones may impact generalization negatively. Inspired by the discovery, we propose a novel training strategy termed Modular Adaptive Training (MAT) to update those modules with their λmax exceeding a dynamic threshold selectively, concentrating the model on learning common features and ignoring those superfluous ones. Unlike most existing training schemes with a complete BP cycle across all network modules, MAT can significantly save computations by its partially-updating strategy and can further improve performance. Experiments show that MAT nearly halves the computational cost of model training and outperforms the accuracy of baselines. © 2024, CC BY.

关键词： Eigenvalues and eigenfunctions

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：