检索结果-内蒙古大学图书馆

Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA)

作者： Mengjie Qin Zheyuan Lin Minhong Wan Chunlong Zhang Jason Gu Te Li Research Center for Intelligent Robotics Research Institute of Interdisciplinary Innovation Zhejiang Lab Hangzhou China Zhejiang Engineering Research Center for Intelligent Robotics Hangzhou China Department of Electrical and Computer Engineering Dalhousie University Halifax

This study focuses on exploring the problem of low-rank tensor completion (LRTC), where the main challenge lies in accurately representing the inherent structures of the underlying data. Typical approaches exhibit either suboptimal performance due to the lack of suitable priors or combine many constraints, resulting in increased model complexity. In this paper, we propose a robust transformed t-SVD by utilizing the discrete cosine transform in conjunction with a unitary transformation matrix. Nonetheless, the t-SVD-based framework exhibits a deficiency in its capacity to flexibly explore diverse correlations across tensor modes. To address this issue, we employ the representation of the tensor as a series of multi-dimensional unfolding tensors, which fully capture the embedded structure in the original data. Moreover, on the basis of this tensor representation, we construct a transformed t-SVD method within the LRTC model and solve it by the alternating-direction method of multiplier (ADMM) approach. Extensive experiments demonstrate that the proposed model outperforms existing state-of-the-art methods.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Pose Graph Optimization over Planar Unit Dual Quaternions: Improved Accuracy with Provably Convergent Riemannian Optimization

Pose Graph Optimization over Planar Unit Dual Quaternions: I...

引用

IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

作者： William D. Warke J. Humberto Ramos Prashant Ganesh Kevin M. Brink Matthew T. Hale Department of Mechanical and Aerospace Engineering University of Florida FL USA Principal Research Engineer EpiSci Science Incorporated Air Force Research Lab Munitions Directorate at Eglin Air Force Base Department of Electrical and Computer Engineering Georgia Institute of Technology Atlanta GA USA

ISBN: (数字)9798350377705

ISBN: (纸本)9798350377712

It is common in pose graph optimization (PGO) algorithms to assume that noise in the translations and rotations of relative pose measurements is uncorrelated. However, existing work shows that in practice these measurements can be highly correlated, which leads to degradation in the accuracy of PGO solutions that rely on this assumption. Therefore, in this paper we develop a novel algorithm derived from a realistic, correlated model of relative pose uncertainty, and we quantify the resulting improvement in the accuracy of the solutions we obtain relative to state-of-the-art PGO algorithms. Our approach utilizes Riemannian optimization on the planar unit dual quaternion (PUDQ) manifold, and we prove that it converges to first-order stationary points of a Lie-theoretic maximum likelihood objective. Then we show experimentally that, compared to state-of-the-art PGO algorithms, this algorithm produces estimation errors that are lower by 10% to 25% across several orders of magnitude of correlated noise levels and graph sizes.

关键词： Accuracy Uncertainty Three-dimensional displays Quaternions Approximation algorithms Robot sensing systems Numerical models Noise measurement Optimization Rotation measurement

来源：评论

学校读者我要写书评

暂无评论

ICSGD-Momentum: SGD Momentum based on Inter-gradient Collision 22

ICSGD-Momentum: SGD Momentum based on Inter-gradient Collisi...

引用

22nd IEEE International Conference on Industrial Informatics, INDIN 2024

作者： Zou, Weidong Cao, Weipeng Xia, Yuanqing Zhong, Bineng Li, Dachuan School of Automation Beijing Institute of Technology Beijing China Shenzhen China Guangxi Key Lab of Multi-source Information Mining & Security Guangxi Normal University Guilin China Research Institute of Trustworthy Autonomous Systems Southern University of Science and Technology Shenzhen China Department of Computer Science and Engineering Southern University of Science and Technology Shenzhen China

ISBN: (纸本)9798331527471

Deep neural networks (DNNs) are widely used in fields like computer vision and natural language processing. A key component of DNN training is the optimizer. SGD-Momentum is popular in many DNN methodologies, such as ResNet and DenseNet, due to its simplicity and effectiveness. However, its slow convergence rate limits its use. To overcome this, we introduce inter-gradient collision into SGD-Momentum, inspired by the elastic collision model in physics. This new method, called ICSGD-Momentum, aims to improve convergence. We provide theoretical proof of convergence and establish a regret bound for ICSGD-Momentum. Experiments on benchmarks including function optimization, CIFAR-100, ImageNet, Penn Treebank, COCO, and YCB-Video show that ICSGD-Momentum accelerates training and enhances the generalization performance of DNNs compared to optimizers like SGD-Momentum, Adam, RAdam, Adabound, and AdaBelief. © 2024 IEEE.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

BinaryAI: Binary Software Composition Analysis via Intelligent Binary Source Code Matching 24

BinaryAI: Binary Software Composition Analysis via Intellige...

引用

44th ACM/IEEE International Conference on Software engineering, ICSE 2024

作者： Jiang, Ling An, Junwen Huang, Huihui Tang, Qiyi Nie, Sen Wu, Shi Zhang, Yuqun Southern University of Science and Technology Shenzhen China Tencent Security Keen Lab Shanghai China Research Institute of Trustworthy Autonomous Systems Southern University of Science and Technology Shenzhen China Southern University of Science and Technol-ogy Department of Computer Science and Engineering Shenzhen China Guangdong Provincial Key Laboratory of Brain-inspired Intelligent Computation China

ISBN: (纸本)9798400702174

While third-party libraries (TPLs) are extensively reused to enhance productivity during software development, they can also introduce potential security risks such as vulnerability propagation. Software composition analysis (SCA), proposed to identify reused TPLs for reducing such risks, has become an essential procedure within modern DevSecOps. As one of the mainstream SCA techniques, binary-to-source SCA identifies the third-party source projects contained in binary files via binary source code matching, which is a major challenge in reverse engineering since binary and source code exhibit substantial disparities after compilation. The existing binary-to-source SCA techniques leverage basic syntactic features that suffer from redundancy and lack robustness in the large-scale TPL dataset, leading to inevitable false positives and compromised recall. To mitigate these limitations, we introduce BinaryAI, a novel binary-to-source SCA technique with two-phase binary source code matching to capture both syntactic and semantic code features. First, BinaryAI trains a transformer-based model to produce function-level embeddings and obtain similar source functions for each binary function accordingly. Then by applying the link-time locality to facilitate function matching, BinaryAI detects the reused TPLs based on the ratio of matched source functions. Our experimental results demonstrate the superior performance of BinaryAI in terms of binary source code matching and the downstream SCA task. Specifically, our embedding model outperforms the state-of-the-art model CodeCMR, i.e., achieving 22.54% recall@l and 0.34 MRR compared with 10.75% and 0.17 respectively. Additionally, BinaryAI outperforms all existing binary-to-source SCA tools in TPL detection, increasing the precision from 73.36% to 85.84% and recall from 59.81% to 64.98% compared with the well-recognized commercial SCA product Black Duck. E-https://*** © 2024 ACM.

关键词： Syntactics

来源：评论

学校读者我要写书评

暂无评论

Expediting Distributed GNN Training with Feature-only Partition and Optimized Communication Planning

Expediting Distributed GNN Training with Feature-only Partit...

引用

IEEE Annual Joint Conference: INFOCOM, IEEE computer and Communications Societies

作者： Bingqian Du Jun Liu Ziyue Luo Chuan Wu Qiankun Zhang Hai Jin National Engineering Research Center for Big Data Technology and System Services Computing Technology and System Lab Cluster and Grid Computing Lab School of Computer Science and Technology Huazhong University of Science and Technology Wuhan China Department of Electrical and Computer Engineering The Ohio State University USA Department of Computer Science The University of Hong Kong Hong Kong School of Cyber Science and Engineering Huazhong University of Science and Technology Wuhan China

ISBN: (数字)9798350383508

ISBN: (纸本)9798350383515

Feature-only partition of large graph data in distributed Graph Neural Network (GNN) training offers advantages over commonly adopted graph structure partition, such as minimal graph preprocessing cost and elimination of cross-worker subgraph sampling burdens. Nonetheless, performance bottleneck of GNN training with feature-only partitions still largely lies in the substantial communication overhead due to cross-worker feature fetching. To reduce the communication overhead and expedite distributed training, we first investigate and answer two key questions on convergence behaviors of GNN model in feature-partition based distribute GNN training: 1) As no worker holds a complete copy of each feature, can gradient exchange among workers compensate for the information loss due to incomplete local features? 2) If the answer to the first question is negative, is feature fetching in every training iteration of the GNN model necessary to ensure model convergence? Based on our theoretical findings on these questions, we derive an optimal communication plan that decides the frequency for feature fetching during the training process, taking into account bandwidth levels among workers and striking a balance between model loss and training time. Extensive evaluation demonstrates consistent results with our theoretical analysis, and the effectiveness of our proposed design.

关键词： Training Time-frequency analysis Costs Computational modeling Distributed databases Bandwidth Graph neural networks

来源：评论

学校读者我要写书评

暂无评论

Cascaded Deep CNN Network for Breast Cancer Screening from Ultrasound Images Using Differential Evolution Algorithm

Cascaded Deep CNN Network for Breast Cancer Screening from U...

引用

Digital Society and Intelligent Systems (DSInS), International Conference on

作者： Wahidur Rahman Md Shadman Sakib Md Solaiman Hosen Mir Mohammad Azad Hani Alshahrani Adel Sulaiman Mana Saleh Al Reshan Asadullah Shaikh Dept. of Computer Science and Engineering Uttara University Uttara Dhaka Bangladesh Dept. of Computer Science and Engineering University of Information Technology & Sciences Dhaka Bangladesh Department of Computer Science Iowa State University USA Department of Computer Science and Engineering Dhaka Bangladesh Department of Computer Science College of Computer Science and Information Systems Najran University Saudi Arabia Emerging Technologies Research Lab (ETRL) College of Computer Science and Information Systems Najran University Saudi Arabia Department of Information Systems College of Computer Science and Information Systems Najran University Saudi Arabia

ISBN: (数字)9798331528829

ISBN: (纸本)9798331528836

Because of poor blood flow and susceptibility to infections, breast cancer can be challenging to heal and, if not treated, might cause major problems, including limb amputation and a lowered quality of life. Though several systems exist to identify breast cancer, few combine machine learning (ML), deep learning (DL), and optimization strategies. This work presents a novel method leveraging complex algorithms to precisely identify breast cancer from ultrasound images. The study utilizes a dataset from a secondary source, categorizing it into three identical classes: normal (Class 0), benign (Class 1), and malignant (Class 2). It extracts features using pre-trained convolutional neural networks (CNNs), optimizes these features using the Differential Evolution Algorithm (DEA), and classifies images using standard machine learning algorithms. To show how well it detects breast cancer, the approach combines seven conventional ML classifiers with ten pre-trained CNN models. Using DenseNet121 + DEA + Xtreme Gradient Boosting (XGB) Classifier, DEA chooses important features derived by CNNs with a high accuracy of 97.76%. However, the proposed method can be very effective for real-time breast cancer screening using ultrasound images.

关键词： Limbs Ultrasonic imaging Machine learning algorithms Feature extraction Breast cancer Real-time systems Classification algorithms Convolutional neural networks Standards Optimization

来源：评论

学校读者我要写书评

暂无评论

Optimizing Antenna Design for 6G Applications Using Machine Learning Techniques

Optimizing Antenna Design for 6G Applications Using Machine ...

引用

Artificial Intelligence and Emerging Technology (Global AI Summit), International Conference on

作者： Praveen Kumar Sharma Vivek Prakash Yadav Aashiwal Nirmal Mukesh Sourabh Jawale Ankit Kumar Research Center for Electrical and Information Technology Seoul National University of Science and Technology Seoul Republic of Korea Department of Electronics and Communication Engineering Independent Researcher: FEAL: Flexible Electronics and Antenna Lab Pilani India Department of Electronics and Communication Engineering Independent Researcher: FEAL: Flexible Electronics and Antenna Lab Mumbai India Department of Computer Science and Engineering Medicaps University Indore India Department of Computer Science and Engineering Seoul National University of Science and Technology Seoul Republic of Korea

ISBN: (数字)9798350379716

ISBN: (纸本)9798350379723

The emergence of 6G networks requires innovative antenna designs that operate at extremely high frequencies. The proposed paper explores the application of advanced machine learning techniques to optimize the design of antennas for 6G applications at 15 GHz. Traditional design methods are often time-consuming and depend on expert knowledge, whereas machine learning can automate and expedite this process. We developed a dataset incorporating critical antenna design parameters such as geometric dimensions and applied algorithms to predict key performance metrics like return loss, gain, and bandwidth. Our results demonstrate that neural networks, in particular, offer superior predictive accuracy, significantly reducing design iteration times. This integration of machine learning into antenna design not only improves performance but also facilitates the development of adaptive antenna systems, meeting the stringent demands of 6G networks. This research highlights the potential of machine learning to revolutionize antenna design, enabling rapid and efficient development of next-generation communication systems.

关键词： 6G mobile communication Technological innovation Machine learning algorithms Communication systems Design methodology Neural networks Machine learning Prediction algorithms Antennas Next generation networking

来源：评论

学校读者我要写书评

暂无评论

Improving Gloss-free Sign Language Translation by Reducing Representation Density

arXiv

引用

arXiv 2024年

作者： Ye, Jinhui Wang, Xing Jiao, Wenxiang Liang, Junwei Xiong, Hui Guangzhou China Department of Computer Science and Engineering HKUST Hong Kong Guangzhou HKUST Fok Ying Tung Research Institute Hong Kong Tencent AI Lab China

Gloss-free sign language translation (SLT) aims to develop well-performing SLT systems with no requirement for the costly gloss annotations, but currently still lags behind gloss-based approaches significantly. In this paper, we identify a representation density problem that could be a bottleneck in restricting the performance of gloss-free SLT. Specifically, the representation density problem describes that the visual representations of semantically distinct sign gestures tend to be closely packed together in feature space, which makes gloss-free methods struggle with distinguishing different sign gestures and suffer from a sharp performance drop. To address the representation density problem, we introduce a simple but effective contrastive learning strategy, namely SignCL, which encourages gloss-free models to learn more discriminative feature representation in a self-supervised manner. Our experiments demonstrate that the proposed SignCL can significantly reduce the representation density and improve performance across various translation frameworks. Specifically, SignCL achieves a significant improvement in BLEU score for the Sign Language Transformer and GFSLT-VLP on the CSL-Daily dataset by 39% and 46%, respectively, without any increase of model parameters. Compared to Sign2GPT, a state-of-the-art method based on large-scale pre-trained vision and language models, SignCL achieves better performance with only 35% of its parameters. Implementation and Checkpoints are available at https://***/JinhuiYE/SignCL. © 2024, CC BY-NC-ND.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

A comparative study of machine learning models with LASSO and SHAP feature selection for breast cancer prediction

引用

Healthcare Analytics 2024年 6卷 100353-100353页

作者： Shaon, Md. Shazzad Hossain Karim, Tasmin Shakil, Md. Shahriar Hasan, Md. Zahid Health Informatics Research Lab Department of Computer Science and Engineering Daffodil International University Daffodil Smart City Dhaka Birulia 1216 Bangladesh

In recent decades, breast cancer has become the most prevalent type of cancer that impacts women in the world, which shows a significant risk to the death rates of women. Early identification of breast cancer might drastically decrease patient mortality and greatly improve the chance of an effective treatment. In modern times, machine learning models have become crucial for classifying cancer and strengthening both the accuracy and efficiency of diagnostic and medical treatment strategies. Therefore, this study is focused on early detection of breast cancer using a variety of machine learning algorithms and desires to identify the most effective feature selection process with an amalgamated dataset. Initially, we evaluated five traditional models and two meta-models on separate datasets. To find the most valuable features, the study used the Least Absolute Shrinkage and Selection Operator (LASSO) as well as SHapley Additive exPlanations (SHAP) selection methods and analyzed them through a wide range of performance regulations. Additionally, we applied these models to the combined dataset and observed that the mergeddataset was significantly beneficial for breast cancer diagnosis. After analyzing the feature selection strategies, it was demonstrated that the majority of models performed more accurately when utilizing SHAP methodologies. Notably, three traditional models and two meta-classifiers obtained an accuracy of 99.82%, demonstrating superior performance compared to state-of-the-art methods. This advancement holds a crucial role as it lays the foundation for refining diagnostic tools and enhancing the progression of medical science in this field. © 2024

关键词： Breast cancer Diagnostic tools Feature selection Machine learning Meta-models SHapley additive exPlanations (SHAP)

来源：评论

学校读者我要写书评

暂无评论

Joint Robotic Aerial Base Station Deployment and Wireless Backhauling in 6G Multi-hop Networks

arXiv

引用

arXiv 2024年

作者： Shang, Wen Liao, Yuan Friderikos, Vasilis Yanikomeroglu, Halim Center of Telecommunication Research King’s College London London United Kingdom Non-Terrestrial Networks Lab Department of Systems and Computer Engineering Carleton University Ottawa Canada

Due to their ability to anchor into tall urban landforms, such as lampposts or street lights, robotic aerial base stations (RABSs) can create a hyper-flexible wireless multi-hop heterogeneous network to meet the forthcoming green, densified, and dynamic network deployment to support, inter alia, high data rates. In this work, we propose a network infrastructure that can concurrently support the wireless backhaul link capacity and access link traffic demand in the millimeter-wave (mmWave) frequency band. The RABSs grasping locations, resource blocks (RBs) assignment, and route flow control are simultaneously optimized to maximize the served traffic demands. Robotic base stations capitalize on the fact that traffic distribution varies considerably across both time and space within a given geographical area. Hence, they are able to relocate to suitable locations, i.e.,’follow’ the traffic demand as it unfolds to increase the overall network efficiency. To tackle the curse of dimensionality of the proposed mixed-integer linear problem, we propose a greedy algorithm to obtain a competitive solution with low computational complexity. Compared to baseline models, which are heterogeneous networks with randomly deployed fixed small cells and pre-allocated RBs for wireless access and backhaul links, a wide set of numerical investigations reveals that robotic base stations could improve the served traffic demand. Specifically, the proposed mode serves at most 65% more traffic demand compared to an equal number of deployed fixed small cells. © 2024, CC BY.

关键词： Antennas

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：