检索结果-内蒙古大学图书馆

TransXNet: Learning Both Global and Local Dynamics With a Dual Dynamic Token Mixer for Visual Recognition

IEEE Transactions on Neural Networks and Learning Systems 2025年第6期36卷 11534-11547页

作者： Meng Lou Shu Zhang Hong-Yu Zhou Sibei Yang Chuan Wu Yizhou Yu School of Computing and Data Science The University of Hong Kong Hong Kong SAR China Artificial Intelligence Laboratory Deepwise Healthcare Beijing China Department of Biomedical Informatics Harvard Medical School Boston CA USA Department of Computer Science The University of Hong Kong Hong Kong SAR China School of Information Science and Technology ShanghaiTech University Shanghai China

Recent studies have integrated convolutions into transformers to introduce inductive bias and improve generalization performance. However, the static nature of conventional convolution prevents it from dynamically adapting to input variations, resulting in a representation discrepancy between convolution and self-attention as self-attention calculates attention matrices dynamically. Furthermore, when stacking token mixers that consist of convolution and self-attention to form a deep network, the static nature of convolution hinders the fusion of features previously generated by self-attention into convolution kernels. These two limitations result in a suboptimal representation capacity of the constructed networks. To find a solution, we propose a lightweight dual dynamic token mixer (D-Mixer) to simultaneously learn global and local dynamics, that is, mechanisms that compute weights for aggregating global contexts and local details in an input-dependent manner. D-Mixer works by applying an efficient global attention module and an input-dependent depthwise convolution separately on evenly split feature segments, endowing the network with strong inductive bias and an enlarged effective receptive field. We use D-Mixer as the basic building block to design TransXNet, a novel hybrid CNN–transformer vision backbone network that delivers compelling performance. In the ImageNet-1K image classification task, TransXNet-T surpasses Swin-T by 0.3% in top-1 accuracy while requiring less than half of the computational cost. Furthermore, TransXNet-S and TransXNet-B exhibit excellent model scalability, achieving top-1 accuracy of 83.8% and 84.6%, respectively, with reasonable computational costs. In addition, our proposed network architecture demonstrates strong generalization capabilities in various dense prediction tasks, outperforming other state-of-the-art networks while having lower computational costs. Code is publicly available at https://***/LMMMEng/TransXNet .

关键词： Mixers Transformers Computational modeling Kernel Computational efficiency Feature extraction Convolutional codes Computer vision Accuracy Stacking

来源：评论

学校读者我要写书评

暂无评论

Arabic Sentiment Analysis of Food Delivery Services Reviews

Arabic Sentiment Analysis of Food Delivery Services Reviews

引用

2023 International Symposium on Networks, Computers and Communications, ISNCC 2023

作者： Mustafa, Dheya Khabour, Safaa M. Shatnawi, Ahmed S. Taqieddin, Eyad The Hashemite University Faculty of Engineering Dept. of Computer Engineering Zarqa13133 Jordan Yarmouk University Faculty of Information Technology and Computer Sciences Dept. of Information Systems Irbid21163 Jordan Jordan University of Science & Technology Dept. of Software Engineering Irbid21110 Jordan School of Computing and Informatics Dept. of Cyber Security Al-Hussein Technical University Amman Jordan

ISBN: (纸本)9798350335590

Customer reviews on online platforms have grown to become an important source of insight into a company's performance. Food delivery services (FDS) companies aim to effectively use customers' feedback to identify areas for improvement of customer satisfaction. Although Arabic is becoming one of the most widely used languages on the Internet, only a few studies have focused on Arabic sentiment analysis to date. The present study conducts an extensive emotion mining and sentiment analysis on FDS-related reviews in Arabic, exploiting natural language processing, and machine learning techniques to extract information that reflects customers' subjective viewpoints, recognize their feelings, and determine polarity in the FDS domain. This work begins with collecting the FDS dataset (Talabat), and then extracting and creating a dialects lexicon for Arabic dialects, followed by walking the reader through detailed steps of cleaning and pre-processing a manually annotated dataset. Finally, we examined classification algorithms including Decision Tree (DT), and Support Vector Machine (SVM). We achieved a maximum accuracy of about 82% using the SVM classifier. © 2023 IEEE.

关键词： Sentiment analysis

来源：评论

学校读者我要写书评

暂无评论

Computation of the exponential function of matrices by a formula without oscillatory integrals on infinite intervals

arXiv

引用

arXiv 2024年

作者： Suzuki, Masato Tanaka, Kenichiro Department of Mathematical Informatics Graduate School of Information Science and Technology University of Tokyo 7-3-1 Hongo Bunkyo-ku Tokyo113-8656 Japan Department of Mathematical and Computing Science School of Computing Institute of Science Tokyo 2-12-1 Ookayama Tokyo Meguro-ku152-8550 Japan

We propose a quadrature-based formula for computing the exponential function of matrices with a non-oscillatory integral on an infinite interval and an oscillatory integral on a finite interval. In the literature, existing quadrature-based formulas are based on the inverse Laplace transform or the Fourier transform. We show these expressions are essentially equivalent in terms of complex integrals and choose the former as a starting point to reduce computational cost. By choosing a simple integral path, we derive an integral expression mentioned above. Then, we can easily apply the double-exponential formula and the Gauss-Legendre formula, which have rigorous error bounds. As numerical experiments show, the proposed formula outperforms the existing formulas when the imaginary parts of the eigenvalues of matrices have large absolute *** Codes 65F60, 65D30 Copyright © 2024, The Authors. All rights reserved.

关键词： Choquet integral

来源：评论

学校读者我要写书评

暂无评论

FineDiffusion: Scaling up Diffusion Models for Fine-grained Image Generation with 10,000 Classes

arXiv

引用

arXiv 2024年

作者： Pan, Ziying Wang, Kun Li, Gang He, Feihong Lai, Yongxuan School of Informatics Xiamen University China Institute of Software Chinese Academy of Sciences China School of Mathematics and Information Engineering Longyan University China Shenyang Institute of Computing Technology of Chinese Academy of Sciences China University of Chinese Academy of Sciences China School of Computer Science and Technology Soochow University China

The class-conditional image generation based on diffusion models is renowned for generating high-quality and diverse images. However, most prior efforts focus on generating images for general categories, e.g., 1000 classes in ImageNet-1k. A more challenging task, large-scale fine-grained image generation, remains the boundary to explore. In this work, we present a parameter-efficient strategy, called FineDiffusion, to fine-tune large pre-trained diffusion models scaling to large-scale fine-grained image generation with 10,000 categories. FineDiffusion significantly accelerates training and reduces storage overhead by only fine-tuning tiered class embedder, bias terms, and normalization layers' parameters. To further improve the image generation quality of fine-grained categories, we propose a novel sampling method for fine-grained image generation, which utilizes superclass-conditioned guidance, specifically tailored for fine-grained categories, to replace the conventional classifier-free guidance sampling. Compared to full fine-tuning, FineDiffusion achieves a remarkable 1.56× training speed-up and requires storing merely 1.77% of the total model parameters, while achieving state-of-the-art FID of 9.776 on image generation of 10,000 classes. Extensive qualitative and quantitative experiments demonstrate the superiority of our method compared to other parameter-efficient fine-tuning methods. The code and more generated results are available at our project website: https://***/. © 2024, CC BY.

关键词： Image enhancement

来源：评论

学校读者我要写书评

暂无评论

Investigation of the separation zone for split-domain light field visualization 28

Investigation of the separation zone for split-domain light ...

引用

2024 28th International Conference on Information technology, IT 2024

作者： Javidi, Kamran Martini, Maria G. Kara, Peter A. School of Computer Science and Mathematics Faculty of Engineering Computing and the Environment Kingston University Wireless Multimedia and Networking Research Group Department of Computer Science Penrhyn Road Campus Kingston upon Thames LondonKT1 2EE United Kingdom Budapest University of Technology and Economics Faculty of Electrical Engineering and Informatics Department of Networked Systems and Services Budapest1111 Hungary

ISBN: (纸本)9798350369618

Light field visualization commonly provides a single content over the entire field of view. However, the angularly-selective nature of the technology enables the simultaneous visualization of different contents at different viewing angles. Yet segmenting the valid viewing area comes with content interference as well in forms of separation zones, resulting invalid viewing domains. In this paper, we introduce a study on the separation zone of split-domain light field visualization. We created a static-content-based scenario in which we split the valid viewing area into two distinct domains, with a separation zone in between them. This was achieved by merging two light field contents in the middle with an instantaneous switch. The resulting visualization on a light field display was captured by a DSLR camera from two viewing distances, and it was compared to the crosstalk effect caused by insufficient angular resolution. © 2024 IEEE.

关键词： Visualization

来源：评论

学校读者我要写书评

暂无评论

Reduced Noise SMOTE in Machine Learning Model: Application in Water Quality Classification with Imbalanced Datasets

Reduced Noise SMOTE in Machine Learning Model: Application i...

引用

International Conference on Artificial Intelligence and Data Sciences (AiDAS)

作者： Norashikin Nasaruddin Nurulkamal Masseran Wan Mohd Razi Idris Ahmad Zia Ul-Saufie Department of Mathematical Sciences Faculty of Science and Technology Universiti Kebangsaan Malaysia Bangi Selangor Malaysia School of Mathematical Sciences College of Computing Informatics and Mathematics Universiti Teknologi Mara (UiTM) Kedah Branch Merbok Kedah Malaysia Department of Earth Science and Environment Faculty of Science and Technology Universiti Kebangsaan Malaysia Bangi Selangor Malaysia School of Mathematical Sciences College of Computing Informatics and Mathematics Universiti Teknologi Mara (UiTM) Shah Alam Selangor Malaysia

ISBN: (数字)9798331528553

ISBN: (纸本)9798331528560

Achieving accurate classification in imbalanced datasets, especially for environmental data such as water quality assessment, is a major challenge for machine learning classifiers. This study introduces the Reduced Noise-Synthetic Minority Oversampling Technique (RN-SMOTE) to address the problems of underrepresentation of minority classes and noise in imbalanced multiclass datasets for Water Quality Classification (WQC). Current state-of-the-art techniques, such as the standard Synthetic Minority Oversampling Technique (SMOTE) and its variants, improve class balance but often need to adequately address noise in synthetic samples. Our research extends RN-SMOTE, previously applied to binary data, to multiclass scenarios. RN-SMOTE improves classification performance by oversampling the minority class and eliminating noisy synthetic instances. We evaluate the effectiveness of RN-SMOTE using three classifiers: Decision Tree (DT), Random Forest (RF), and Extreme Gradient Boosting (XGBoost). The experimental results reveal that RN-SMOTE significantly improves classification accuracy and sensitivity. For instance, the RF classifier with RN-SMOTE achieved an accuracy of 71.17% and a sensitivity of 75.24%, 69.23% and 72.14% for the clean, slightly polluted and polluted classes, respectively, outperforming the original dataset and traditional SMOTE techniques. However, RN-SMOTE did not outperform the traditional SMOTE for DT and the XGBoost model. Applying RN-SMOTE to multiclass water quality data extends its utility and advances unbalanced classification in environmental science.

关键词： Sensitivity Accuracy Pollution Noise Training data Water quality Data models Reliability Standards Classification tree analysis

来源：评论

学校读者我要写书评

暂无评论

SketchDream: Sketch-based Text-to-3D Generation and Editing

arXiv

引用

arXiv 2024年

作者： Liu, Feng-Lin Fu, Hongbo Lai, Yu-Kun Gao, Lin SCM City University of Hong Kong EMIA HKUST China School of Computer Science and Informatics Cardiff University United Kingdom The Beijing Key Laboratory of Mobile Computing and Pervasive Device Institute of Computing Technology Chinese Academy of Sciences University of Chinese Academy of Sciences China

Existing text-based 3D generation methods generate attractive results but lack detailed geometry control. Sketches, known for their conciseness and expressiveness, have contributed to intuitive 3D modeling but are confined to producing texture-less mesh models within predefined categories. Integrating sketch and text simultaneously for 3D generation promises enhanced control over geometry and appearance but faces challenges from 2D-to-3D translation ambiguity and multi-modal condition integration. Moreover, further editing of 3D models in arbitrary views will give users more freedom to customize their models. However, it is difficult to achieve high generation quality, preserve unedited regions, and manage proper interactions between shape components. To solve the above issues, we propose a text-driven 3D content generation and editing method, SketchDream, which supports NeRF generation from given hand-drawn sketches and achieves free-view sketch-based local editing. To tackle the 2D-to-3D ambiguity challenge, we introduce a sketch-based multi-view image generation diffusion model, which leverages depth guidance to establish spatial correspondence. A 3D ControlNet with a 3D attention module is utilized to control multi-view images and ensure their 3D consistency. To support local editing, we further propose a coarse-to-fine editing approach: the coarse phase analyzes component interactions and provides 3D masks to label edited regions, while the fine stage generates realistic results with refined details by local enhancement. Extensive experiments validate that our method generates higher-quality results compared with a combination of 2D ControlNet and image-to-3D generation techniques and achieves detailed control compared with existing diffusion-based 3D editing approaches. Copyright © 2024, The Authors. All rights reserved.

关键词： 3D modeling

来源：评论

学校读者我要写书评

暂无评论

FROM INTENTION TO IMPLEMENTATION: AUTOMATING BIOMEDICAL RESEARCH VIA LLMS

arXiv

引用

arXiv 2024年

作者： Luo, Yi Shi, Linghang Li, Yihao Zhuang, Aobo Gong, Yeyun Liu, Ling Chen, Lin School of Informatics Xiamen University China National Institute for Data Science in Health and Medicine Xiamen University China School of Medicine Xiamen University China Microsoft Research Asia China College of Computing Georgia Institute of Technology United States

Conventional biomedical research is increasingly labor-intensive due to the exponential growth of scientific literature and datasets. Artificial intelligence (AI), particularly Large Language Models (LLMs), has the potential to revolutionize this process by automating various steps. Still, significant challenges remain, including the need for multidisciplinary expertise, logicality of experimental design, and performance measurements. This paper introduces BioResearcher, the first end-to-end automated system designed to streamline the entire biomedical research process involving dry lab experiments. BioResearcher employs a modular multi-agent architecture, integrating specialized agents for search, literature processing, experimental design, and programming. By decomposing complex tasks into logically related sub-tasks and utilizing a hierarchical learning approach, BioResearcher effectively addresses the challenges of multidisciplinary requirements and logical complexity. Furthermore, BioResearcher incorporates an LLM-based reviewer for in-process quality control and introduces novel evaluation metrics to assess the quality and automation of experimental protocols. BioResearcher successfully achieves an average execution success rate of 63.07% across eight previously unmet research objectives. The generated protocols averagely outperform typical agent systems by 22.0% on five quality metrics. The system demonstrates significant potential to reduce researchers’ workloads and accelerate biomedical discoveries, paving the way for future innovations in automated research systems. Copyright © 2024, The Authors. All rights reserved.

关键词： Hierarchical systems

来源：评论

学校读者我要写书评

暂无评论

Neural Manifold Learning Based on 40 nm Dual-Mode PCM Compute-in-Memory Chip with Hardware Adaptive Drift Compensation

Neural Manifold Learning Based on 40 nm Dual-Mode PCM Comput...

引用

2024 IEEE International Electron Devices Meeting, IEDM 2024

作者： Yan, Longhao Li, Yuqi Li, Xi Pan, Zelun Wang, Zeyu Wang, Xile Wang, Bowen Zhan, Zhe Tang, Xiyuan Tao, Yaoyu Ge, Woo-Ping Song, Zhitang Huang, Ru Yang, Yuchao Beijing Advanced Innovation Center for Integrated Circuits School of Integrate Circuits Peking University Beijing China Beijing China Center for Brain Inspired Chips Institute for Artificial Intelligence Peking University Beijing China Guangdong Provincial Key Laboratory of In-Memory Computing Chips School of Electronic and Computer Engineering Peking University Shenzhen China State Key Laboratory of Functional Materials for Informatics Shanghai Institute of Microsystem and Information Technology Chinese Academy of Sciences China

ISBN: (纸本)9798350365429

Neural manifold learning (NML), as a significant research topic in the field of neuroscience, plays a vital role in the realization of intelligent brain-computer interface (BCI) at the edge. However, traditional edge computing platforms fall short in meeting the demands for the huge computing power needed to process massive neural data. Here, we design a 288Kb dual-mode phase change memory (PCM) compute-in-memory chip in 40nm technology and utilize its two cores for two fundamental kernels in NML, vector matrix multiplication (VMM) and true random number generation, and develop a PCM-based NML system for the first time. In particular, in order to mitigate the performance degradation caused by PCM conductance drift, we propose a novel adaptive-PCM-drift-compensation scheme that significantly improves the accuracy of VMM results. Compared with CPU platforms running the same tasks, the resulting chip improves the computing energy efficiency and throughput by 476× and 264×, respectively. © 2024 IEEE.

关键词： Random number generation

来源：评论

学校读者我要写书评

暂无评论

FedClust: Tackling Data Heterogeneity in Federated Learning through Weight-Driven Client Clustering

arXiv

引用

arXiv 2024年

作者： Islam, Md Sirajul Yuan, Xu Javaherian, Simin Chen, Li Xu, Fei Tzeng, Nian-Feng School of Computing and Informatics University of Louisiana Lafayette LafayetteLA United States Department of Computer and Information Sciences University of Delaware NewarkDE United States School of Computer Science and Technology East China Normal University Shanghai China

Federated learning (FL) is an emerging distributed machine learning paradigm that enables collaborative training of machine learning models over decentralized devices without exposing their local data. One of the major challenges in FL is the presence of uneven data distributions across client devices, violating the well-known assumption of independent-and-identically-distributed (IID) training samples in conventional machine learning. To address the performance degradation issue incurred by such data heterogeneity, clustered federated learning (CFL) shows its promise by grouping clients into separate learning clusters based on the similarity of their local data distributions. However, state-of-the-art CFL approaches require a large number of communication rounds to learn the distribution similarities during training until the formation of clusters is stabilized. Moreover, some of these algorithms heavily rely on a predefined number of clusters, thus limiting their flexibility and adaptability. In this paper, we propose FedClust, a novel approach for CFL that leverages the correlation between local model weights and the data distribution of clients. FedClust groups clients into clusters in a one-shot manner by measuring the similarity degrees among clients based on the strategically selected partial weights of locally trained models. We conduct extensive experiments on four benchmark datasets with different non-IID data settings. Experimental results demonstrate that FedClust achieves higher model accuracy up to ∼45% as well as faster convergence with a significantly reduced communication cost up to 2.7× compared to its state-of-the-art counterparts. Copyright © 2024, The Authors. All rights reserved.

关键词： Machine 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：