检索结果-内蒙古大学图书馆

arXiv 2024年

作者： Li, Dachong Li, Li Chen, Zhuangzhuang Li, Jianqiang College of Computer Science and Software Engineering Shenzhen University Shenzhen China School of Mathematical Sciences Shenzhen University Shenzhen China National Engineering Laboratory for Big Data System Computing Technology China

Large kernels make standard convolutional neural networks (CNNs) great again over transformer architectures in various vision tasks. Nonetheless, recent studies meticulously designed around increasing kernel size have shown diminishing returns or stagnation in performance. Thus, the hidden factors of large kernel convolution that affect model performance remain unexplored. In this paper, we reveal that the key hidden factors of large kernels can be summarized as two separate components: extracting features at a certain granularity and fusing features by multiple pathways. To this end, we leverage the multi-path long-distance sparse dependency relationship to enhance feature utilization via the proposed Shiftwise (SW) convolution operator with a pure CNN architecture. In a wide range of vision tasks such as classification, segmentation, and detection, SW surpasses state-of-the-art transformers and CNN architectures, including SLaK and UniRepLKNet. More importantly, our experiments demonstrate that 3 × 3 convolutions can replace large convolutions in existing large kernel CNNs to achieve comparable effects, which may inspire follow-up works. Code and all the models at https://***/lidc54/shift-wiseConv. © 2024, CC BY-NC-SA.

关键词： Convolution

来源：评论

学校读者我要写书评

暂无评论

A Platform Independent Model Design and Adaptation for Edge Intelligence

A Platform Independent Model Design and Adaptation for Edge ...

引用

International Conference on Automation and computing (ICAC)

作者： Chaolong Zhang Jiliu Zhou Jia He Zhijie Xu Jin Jin Chao Kong Yuanping Xu Benjun Guo Qiuyan Gai School of Software Engineering Chengdu University of Information Technology Chengdu China School of Computer Science Chengdu University of Information Technology Chengdu China School of Computing & Engineering University of Huddersfield Huddersfield UK

ISBN: (数字)9798350360882

ISBN: (纸本)9798350360899

Along with the rapid developments in deep learning and edge computing technologies, deploying models on mobile devices is a trend for modern applications. However, the resource-constrained and heterogeneous edge devices fail to cope with complicated model inference. To tackle this challenge, this study explores the platform independent model design based on Open Neural Network Exchange. Then we quantize the existing heavy deep learning models into lightweight ones by applying low-precision arithmetic and partitioning the computational graph into sub-graphs for parallel execution and hardware-based acceleration on heterogeneous systems. This study evaluates the efficiency and effectiveness among different quantization settings and analyses the impacts for model acceleration. The experiments show that the quantized models with uint8 type computation perform faster inference while still preserve high performance. This result suggests a novel solution for the research on edge intelligence.

关键词： Deep learning Adaptation models Quantization (signal) Automation Computational modeling Neural networks Market research Mobile handsets Edge computing Arithmetic

来源：评论

学校读者我要写书评

暂无评论

Deep spatio-temporal convolutional neural network for city traffic flow prediction 2

Deep spatio-temporal convolutional neural network for city t...

引用

2nd International Conference on computing and Data Science, CDS 2021

作者： Zhou, Zhiyuan Qin, Yanjun Luo, Haiyong School of Engineering Jilin University Changchun China School of Software Engineering University of Posts and Telecommunications Beijing China Research Centre of Pervasive Computing Institute of Computing Technology Chinese Academy of Sciences Beijing China

ISBN: (纸本)9781665404280

Forecasting transportation flow is of vital significance for relieving traffic congestion and improving public safety. However, it is very challenging to achieve it precisely because many factors such as weather condition, traffic control and big celebration events can lay great influence on it. To better fulfill this challenging task, we propose a deep-learning-based approach called Spatio-Temporal Convolutional Neural Network. We first model three temporal properties of transportation flow (closeness, period, trend). Each property is assigned with a convolutional neural network, each of which models the corresponding property of public traffic. This model also fuses the aggregation of the output of the three properties with external elements, for example weather condition and some big events, to gain a better performance in citywide traffic flow prediction. Experiments on Beijing taxi flow and the New York city bike flow show that our ST-CNN model outperforms many well-known passenger flow prediction methods. © 2021 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Learning with Open-world Noisy Data via Class-independent Margin in Dual Representation Space 39

Learning with Open-world Noisy Data via Class-independent Ma...

引用

39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025

作者： Pan, Linchao Gao, Can Zhou, Jie Wang, Jinbao College of Computer Science and Software Engineering Shenzhen University China Guangdong Provincial Key Laboratory of Intelligent Information Processing China National Engineering Laboratory for Big Data System Computing Technology Shenzhen University China

ISBN: (纸本)157735897X

Learning with Noisy Labels (LNL) aims to improve the model generalization when facing data with noisy labels, and existing methods generally assume that noisy labels come from known classes, called closed-set noise. However, in real-world scenarios, noisy labels from similar unknown classes, i.e., open-set noise, may occur during the training and inference stage. Such open-world noisy labels may significantly impact the performance of LNL methods. In this study, we propose a novel dual-space joint learning method to robustly handle the open-world noise. To mitigate model overfitting on closed-set and open-set noises, a dual representation space is constructed by two networks. One is a projection network that learns shared representations in the prototype space, while the other is a One-Vs-All (OVA) network that makes predictions using unique semantic representations in the class-independent space. Then, bi-level contrastive learning and consistency regularization are introduced in two spaces to enhance the detection capability for data with unknown classes. To benefit from the memorization effects across different types of samples, class-independent margin criteria are designed for sample identification, which selects clean samples, weights closed-set noise, and filters open-set noise effectively. Extensive experiments demonstrate that our method outperforms the state-of-the-art methods and achieves an average accuracy improvement of 4.55% and an AUROC improvement of 6.17% on CIFAR80N. Copyright © 2025, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Distinguishability-based genuine nonlocality with genuine multipartite entanglement

引用

Physical Review A 2023年第2期108卷 022405-022405页

作者： Zong-Xing Xiong Mao-Sheng Li Zhu-Jun Zheng Lvzhou Li Institute of Quantum Computing and Software School of Computer Science and Engineering Sun Yat-Sen University Guangzhou 510006 China School of Mathematics South China University of Technology Guangzhou 510641 China

A set of orthogonal multipartite quantum states is said to be distinguishability-based genuinely nonlocal (also genuinely nonlocal, for abbreviation) if the states are locally indistinguishable across any bipartition of the subsystems. This form of multipartite nonlocality, although more naturally arising than the recently popular “strong nonlocality” in the context of local distinguishability, receives much less attention. In this work, we study the distinguishability-based genuine nonlocality of a typical type of genuine multipartite entangled states—the d-dimensional Greenberger-Horne-Zeilinger (GHZ) states, featuring systems with local dimension not limited to two. In the three-partite case, we find the existence of small genuinely nonlocal sets consisting of these states: we show that the cardinality can at least scale down to linear in the local dimension d, with the linear factor l=1. Specifically, the method we use is semidefinite programming and the GHZ states to construct these sets are special ones which we call “GHZ lattices”. This result might arguably suggest a significant gap between the strength of strong nonlocality and the distinguishability-based genuine nonlocality. Moreover, we put forward the notion of (s,n)-threshold distinguishability and, utilizing a similar method, we successfully construct (2,3)-threshold sets consisting of GHZ states in three-partite systems.

关键词： Nonlocality Quantum correlations in quantum information Quantum entanglement

来源：评论

学校读者我要写书评

暂无评论

Learning Multitask Commonness and Uniqueness for Multimodal Sarcasm Detection and Sentiment Analysis in Conversation

IEEE Transactions on Artificial Intelligence

引用

IEEE Transactions on Artificial Intelligence 2024年第3期5卷 1349-1361页

作者： Zhang, Yazhou Yu, Yang Zhao, Dongming Li, Zuhe Wang, Bo Hou, Yuexian Tiwari, Prayag Qin, Jing Centre for Smart Health School of Nursing The Hong Kong Polytechnic University Hong Kong Artificial Intelligence Laboratory China Mobile Communication Group Tianjin Company Ltd. Tianjin300020 China College of Intelligence and Computing Tianjin University Tianjin300072 China Zhengzhou University of Light Industry Zhengzhou450002 China Software Engineering College Zhengzhou University of Light Industry Zhengzhou450002 China College of Computer and Communication Engineering Zhengzhou University of Light Industry Zhengzhou450002 China School of Information Technology Halmstad University Halmstad301 18 Sweden

Sarcasm is a form of figurative language device to express human inner feelings, where the author writes the positive sentence on surface form, while he/she actually expresses negative sentiment, vice versa. Sentiment, thus, comes into sight, and is closely related with sarcasm, leading to the recent popularity of multimodal sarcasm and sentiment joint detection in conversation (dialogue). The key challenges involve multimodal fusion and multitask interaction. Most of the existing studies have focused on building multimodal fused representation, while the commonness and uniqueness across related tasks has not received attention. To fill this gap, we propose a multimodal multitask interaction learning framework, termed MIL, for joint detection of sarcasm and sentiment. Specifically, a cross-modal target attention mechanism is proposed to automatically learn the alignment between texts and images/speeches. In addition, a multimodal interaction learning paradigm consisting of a dual-gating network, three separate fully connected layers that simultaneously capture the commonness and uniqueness. Comprehensive experiments on two benchmarking datasets (i.e., Memotion and MUStARD) show the effectiveness of the proposedmodel over state-of-the-art baselines with a significant improvement of 1.9%, 2.4% in terms of F1. Impact Statement-Human language is multi-affective in nature, where sentiment and sarcasm are often coupled with each *** propose a state-of-the-art multi-task learning (MTL) paradigm to jointly study them, which is of great significance for deepening the understanding of multi-affective analysis and emotional intelligence. The recent works still adopted the hard parameter sharing based MTL paradigm, which mainly focused on the commonness across tasks. Our work is quite different from them, where we make the first attempt to design a soft parameter sharing MTL paradigm, aiming to take the uniqueness of individual task into *** argue that both th

关键词： Sentiment analysis

来源：评论

学校读者我要写书评

暂无评论

Learning with Open-world Noisy Data via Class-independent Margin in Dual Representation Space

arXiv

引用

arXiv 2025年

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

A Grey Wolf Optimization Algorithm with Triangular Community and Crossover Operator for Community Discovery 7

A Grey Wolf Optimization Algorithm with Triangular Community...

引用

7th International Conference on Systems and Informatics, ICSAI 2021

作者： Kang, Yan Huang, Xin Xu, Zhongming Yang, Xuekun Li, Xinyan Software College. Software Engineering of Yunnan University Kunming Data Science and Computing Lab Yunnan China School of Business Travel School of Business and Tourism Management of Yunnan University Kunming Data Science and Computing Lab Yunnan China

ISBN: (纸本)9781665426244

Community discovery under complex networks is a hot discussion issue in network science research. It is very necessary to find a good community structure to study complex networks. At present, many evolutionary algorithms are used for community discovery. However, prior knowledge is not considered for community detection, and full consideration of the network topology can further improve community discovery performance. Therefore, we propose a new algorithm TM-GWO to optimize community discovery. TM-GWO is based on the gray wolf optimization algorithm. It designs a new initialization method and migration-based crossover operator to realize community discovery. The experimental results show that TM-GWO is better than the current Multi-objective evolutionary algorithm. © 2021 IEEE.

关键词： Multiobjective optimization

来源：评论

学校读者我要写书评

暂无评论

Big-Moe: Bypassing Isolated Gating For Generalized Multimodal Face Anti-Spoofing

Big-Moe: Bypassing Isolated Gating For Generalized Multimoda...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Yingjie Ma Zitong Yu Xun Lin Weicheng Xie Linlin Shen College of Computer Science and Software Engineering Shenzhen University Great Bay University National Engineering Laboratory for Big Data System Computing Technology Shenzhen University Guangdong Provincial Key Laboratory of Intelligent Information Processing

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

In the domain of facial recognition security, multimodal Face Anti-Spoofing (FAS) is essential for countering presentation attacks. However, existing technologies encounter challenges due to modality biases and imbalances, as well as domain shifts. Our research introduces a Mixture of Experts (MoE) model to address these issues effectively. We identified three limitations in traditional MoE approaches to multimodal FAS: (1) Coarse-grained experts’ inability to capture nuanced spoofing indicators; (2) Gated networks’ susceptibility to input noise affecting decision-making; (3) MoE’s sensitivity to prompt tokens leading to overfitting with conventional learning methods. To mitigate these, we propose the Bypass Isolated Gating MoE (BIG-MoE) framework, featuring: (1) Fine-grained experts for enhanced detection of subtle spoofing cues; (2) An isolation gating mechanism to counteract input noise; (3) A novel differential convolutional prompt bypass enriching the gating network with critical local features, thereby improving perceptual capabilities. Extensive experiments on four benchmark datasets demonstrate significant generalization performance improvement in multimodal FAS task. The code is released at https://***/murInJ/BIG-MoE.

关键词： Convolutional codes Sensitivity Convolution Face recognition Noise Logic gates Feature extraction Security Speech processing Overfitting

来源：评论

学校读者我要写书评

暂无评论

A Plugin for Kotlin based Android Apps to Detect Security Breaches through Dataflow

A Plugin for Kotlin based Android Apps to Detect Security Br...

引用

IEEE Annual International Computer software and Applications Conference (COMPSAC)

作者： Md Arabin Islam Talukder Sumaiya Farzana Mishu Hossain Shahriar ABM Kamrul Islam Riad Fan Wu Akond Rahman College of Computing and Software Engineering Kennesaw State University Marietta GA USA Department of Computer Science Tuskegee University Tuskegee Alabama USA Department of Computer Science and Software Engineering Auburn University Auburn Alabama USA

Android developers have already adopted Kotlin as their preferred language. Kotlin is more well-accepted than Java because of its simplicity, readability, and new features like scope functions, extension functions, null safety, etc. Google has introduced several new libraries such as RoomDb, Jetpack-Compose, as well as dependency injection frameworks like Dagger-Hilt to enhance Android Development. The question is how secure these libraries are? Are there tools available to analyze new Libraries? DroidPatrol [1] is our existing static analysis plugin that works on Android Apps written in Java. To support Kotlin code analysis we were continuously working on upgrading our plugin. Finally, we released DroidPatrol 2.0 [2] which is compatible to perform static analysis of Android apps developed in Kotlin. It also works on Java-based Android app developed in Java. In this latest edition, we restructured the architecture of the plugin to optimize its efficiency. We also found a vulnerability in RoomDb. The latest version is developed in Kotlin, and we used IntelliJ Idea. Lastly, version 2.0 is independent of Android Studio editions meaning that any version of Android Studio is compatible.

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：