检索结果-内蒙古大学图书馆

International Symposium on Chinese Spoken Language Processing

作者： Ruofan Yan Shu Peng Zhige Chen Zhi-An Huang Rui Liu Kay Chen Tan Jibin Wu Department of Computing Hong Kong Polytechnic University Hong Kong City University (Dong Guan) Department of Data Science and Artificial Intelligence Hong Kong Polytechnic University

ISBN: (数字)9798331516826

ISBN: (纸本)9798331516833

Individuals with hearing impairments often struggle to isolate and focus on a single speaker in multi-speaker environments. Neuroscience research has uncovered distinct patterns of brain activity associated with auditory attention that can be detected using electroencephalography (EEG) measurements. Existing deep learning methods developed for AAD encounter challenges in extracting effective spatial-temporal features and handling the data scarcity issues. To address these issues, this work proposes a novel neural network architecture named ST-AADNet, which integrates a convolutional neural network and a long short-term memory network to extract useful spatial-temporal features from EEG signals. Furthermore, we introduce a series of data augmentation methods tailored to enhance the model's generalization capacity. Experimental studies on both audio-only and audio-video datasets demonstrate the superior performance of the proposed methods, securing the second position in The First Chinese Auditory Attention Decoding Challenge.

关键词： Accuracy Feature extraction data augmentation Brain modeling Electroencephalography Decoding Spatiotemporal phenomena Convolutional neural networks Long short term memory Overfitting

来源：评论

学校读者我要写书评

暂无评论

CET2: Modelling Topic Transitions for Coherent and Engaging Knowledge-Grounded Conversations

arXiv

引用

arXiv 2024年

作者： Xu, Lin Zhou, Qixian Fu, Jinlan Ng, See-Kiong The Department of Computer Science School of Computing National University of Singapore Singapore Institute of Data Science National University of Singapore 117602. Qixian Zhou is with Bytedance Shenzhen China

Knowledge-grounded dialogue systems aim to generate coherent and engaging responses based on the dialogue contexts and selected external knowledge. Previous knowledge selection methods tend to rely too heavily on the dialogue contexts or over-emphasize the new information in the selected knowledge, resulting in the selection of repetitious or incongruous knowledge and further generating repetitive or incoherent responses, as the generation of the response depends on the chosen knowledge. To address these shortcomings, we introduce a Coherent and Engaging Topic Transition (CET2) framework to model topic transitions for selecting knowledge that is coherent to the context of the conversations while providing adequate knowledge diversity for topic development. Our CET2 framework considers multiple factors for knowledge selection, including valid transition logic from dialogue contexts to the following topics and systematic comparisons between available knowledge candidates. Extensive experiments on two public benchmarks demonstrate the superiority and the better generalization ability of CET2 on knowledge selection. This is due to our well-designed transition features and comparative knowledge selection strategy, which are more transferable to conversations about unseen topics. Analysis of fine-grained knowledge selection accuracy also shows that CET2 can better balance topic entailment (contextual coherence) and development (knowledge diversity) in dialogue than existing approaches. © 2024, CC BY-NC-ND.

关键词： Speech processing

来源：评论

学校读者我要写书评

暂无评论

Singe Image Dehazing using Deep Neural Networks

Singe Image Dehazing using Deep Neural Networks

引用

Advances in computing, Communication and Applied Informatics (ACCAI), International Conference on

作者： Nisarg Doshi Sagar Bhavsar Rajeswari D Srinivasan R Department of Data Science and Business Systems School of Computing College of Engineering and Technology SRM Institute of Science and Technology Kattankulathur Tamil Nadu India Department of Computing Technologies School of Computing College of Engineering and Technology SRM Institute of Science and Technology Kattankulathur Tamil Nadu India

In dusty and hazzy circumstances, pictures caputured tend to be unclear. To tackle this challenge researchers have founded different image-dehazing approaches. Lately, high quality pictures which could be used to extract maximal data from autonomous systems are in great demand. In order to minimize the haze from the picture, this research study uses various deep learning (also known as DL) architectures to identify the information recovered and extract key details from the image. The research paper focuses on utilizing deep learning techniques to remove haze from clear *** first phase of the suggested architecture involves applying multiple pre-processing methods, including estimating the airlight, implementing contextual regularization and boundary constraints, as well as contrast enhancement techniques (CS, CLAHE, HE) and Image Fusion. The subsequent goal of this study is to pinpoint the optimal deep learning model that can produce clear images based on the resulting dehazed images. The quality of the dehazed images is evaluated by comparing them with clear images using metrics such as PSNR and SSIM. Experimental findings shows that Alex Net upon the implementation of pre-processing beats other state of the art techniques with regard to PSNR value and holds its own w.r.t SSIM values.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Facial Emotion Recognition using YOLO based Deep Learning Classifier 1

Facial Emotion Recognition using YOLO based Deep Learning Cl...

引用

1st International Conference on Electronics, Communication and Signal Processing, ICECSP 2024

作者： Vanamoju, Sri Vaibhav Mohan Dev Vineetha, Moturi Veda Tekchandani, Hitesh Joshi, Pallavi Shukla, Praveen Kumar Khanna, Anita Electronics and Communication Koneru Lakshmaiah Education Foundation Hyderabad India Artificial Intelligence and Data Science Koneru Lakshmaiah Education Foundation Hyderabad India Amrita Vishwa Vidyapeetham Mysuru Department of Computer Science Mysuru India School of Computing and Intelligent Systems Manipal University Jaipur Jaipur India Electrical Engineering Guru Ghasidas Vishwavidyalaya Bilaspur Bilaspur India

ISBN: (纸本)9798350364590

Understanding facial expressions is key to successful communication. It's like decoding a secret language that helps us know if someone is happy, sad, angry, surprised, or something else altogether. Harnessing facial emotion recognition (FER) in machines holds the potential to revolutionize human-computer interaction, improve customer service experiences and provide invaluable support in medical diagnosis. Machines can comprehend human's needs more effectively by reading the emotions from their facial expression and therefore this results in a more natural conversation between them and the machine. Various machine and deep learning methodologies have been implemented for FER. In this research work, the authors propose the state-of-the-art YOLO (You Only Look Once) based deep learning classifier for FER. The YOLO algorithm is a single-shot approach that efficiently classifies objects in a single pass. It achieves this by using a single neural network to predict bounding boxes and target class probabilities by utilizing the entire image as input. YOLOv8 is a recent iteration of the YOLO object detection algorithm is popular for its speed and accuracy which offers improvements over prior versions, making it a suitable choice for real-time intended task of FER. With the proposed classifier, the authors achieved test accuracy of 72.56% on the FER-2013 dataset in a seven-class FER task. © 2024 IEEE.

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

Oil Spill Detection Using 2D Convolution Neural Network and Generative Adversarial Network

Oil Spill Detection Using 2D Convolution Neural Network and ...

引用

Expert Clouds and Applications (ICOECA), International Conference on

作者： R. Madhura S. Nikkath Bushra A. Syed Ismail Haleema PK Department of Computing Technologies SRM Institute of Science and Technology Kattankulathur Chennai Tamil Nadu India Department of Data Science and Business Systems SRM Institute of Science and Technology Kattankulathur Chennai Tamil Nadu India School of Computing and Engineering University of West London RAK Branch Campus UAE

ISBN: (数字)9798350385793

ISBN: (纸本)9798350385809

Finding oil spills in the ocean is one of the most crucial tasks in preserving our ecosystem. Using satellite or aerial photographs as input to a deep learning model that makes use of 2D CNN (Convolution Neural Network) and GAN (Generative Adversarial Network). Deep learning assists in automatically detecting oil spills in the ocean. A strong model that may be applied to volumetric data is 2D CNN. To protect the environment and its biodiversity, it is critical to identify and minimize oil spills in delicate maritime environments. Conventional techniques for tracking oil spills in satellite imagery sometimes depend on labor-intensive, human error-prone manual inspection. When deep learning models are taught with a lot of data, they perform well. The Gathering of datasets, however the situation proved to be challenging. As a result, employing Generative Adversarial Networks (GAN) to produce new photos by utilizing preexisting ones. As a result, by giving the model a suitable number of photos to use for training and testing, it improves accuracy.

关键词： Deep learning Training Convolution Oils Biological system modeling Oceans Neural networks

来源：评论

学校读者我要写书评

暂无评论

Multi-layer network embedding on scc-based network with motif

引用

Digital Communications and Networks 2024年第3期10卷 546-556页

作者： Lu Sun Xiaona Li Mingyue Zhang Liangtian Wan Yun Lin Xianpeng Wang Gang Xu Department of Communication Engineering Institute of Information Science TechnologyDalian Maritime UniversityDalian116026China Key Laboratory of Big Data Intelligent Computing Chongqing University of Posts and TelecommunicationsChongqing400065China Key Laboratory for Ubiquitous Network and Service Software of Liaoning Province DUT School of Software Technology&DUT-RU International School of Information Science and EngineeringDalian University of TechnologyDalian 116620China College of Information and Communication Engineering Harbin Engineering UniversityHarbin 150001China School of Information and Communication Engineering Hainan UniversityHaikou 570228China State Key Laboratory of Millimeter Waves School of Information Science and EngineeringSoutheast UniversityNanjing 210096China

Interconnection of all things challenges the traditional communication methods,and Semantic Communication and computing(SCC)will become new *** is a challenging task to accurately detect,extract,and represent semantic information in the research of SCC-based *** previous research,researchers usually use convolution to extract the feature information of a graph and perform the corresponding task of node ***,the content of semantic information is quite *** graph convolutional neural networks provide an effective solution for node classification tasks,due to their limitations in representing multiple relational patterns and not recognizing and analyzing higher-order local structures,the extracted feature information is subject to varying degrees of ***,this paper extends from a single-layer topology network to a multi-layer heterogeneous topology *** Bidirectional Encoder Representations from Transformers(BERT)training word vector is introduced to extract the semantic features in the network,and the existing graph neural network is improved by combining the higher-order local feature module of the network model representation network.A multi-layer network embedding algorithm on SCC-based networks with motifs is proposed to complete the task of end-to-end node *** verify the effectiveness of the algorithm on a real multi-layer heterogeneous network.

关键词： Semantic communication and computing Multi-layer network Graph neural network Motif

来源：评论

学校读者我要写书评

暂无评论

Cyberattacks against Direct Load Control of Residential Electric Water Heaters in Smart Grids

Cyberattacks against Direct Load Control of Residential Elec...

引用

2022 IEEE Power and Energy Society Innovative Smart Grid Technologies Conference, ISGT 2022

作者： Youssef, El-Nasser S. Labeau, Fabrice Kassouf, Marthe Alarie, Stéphane McGill University Department of Electrical and Computer Engineering MontrealQC Canada Hydro-Quebec's Research Institute Data Science and High Performance Computing Unit VarennesQC Canada

ISBN: (数字)9781665437752

ISBN: (纸本)9781665437752

The invasive deployment of digital technologies in power grids increases their exposure to cyber threats. In this paper, we study a type of load-altering cyberattack that targets peak shaving programs wherein a utility directly controls the loads of domestic electric water heaters. We consider a cyberattack scenario in which an adversary gains access to the advanced metering infrastructure and sends malicious activation commands to a large number of water heaters forcing them to simultaneously switch on in the midst of a peak shaving period, thus causing a sudden, large power demand that could lead to a system-wide blackout. We develop a simulation environment to simulate the loads of a population of electric water heaters under different operating conditions as well as under this type of cyberattack. Then, we propose a cybersecurity-by-design approach to mitigate the risk of such cyberattack. The efficacy of the proposed solution is analyzed using software experiments. © 2022 IEEE.

关键词： Cybersecurity

来源：评论

学校读者我要写书评

暂无评论

GHVC-Net: Hypervolume Contribution Approximation Based on Graph Neural Network

GHVC-Net: Hypervolume Contribution Approximation Based on Gr...

引用

IEEE International Conference on Systems, Man and Cybernetics

作者： Guotong Wu Yang Nan Ke Shang Hisao Ishibuchi Department of Computer Science and Engineering Southern University of Science and Technology Shenzhen China National Engineering Laboratory for Big Data System Computing Technology Shenzhen University Shenzhen China

ISBN: (数字)9781665410205

ISBN: (纸本)9781665410212

This paper proposes a framework called GHVC-Net that uses the graph neural network (GNN) model to approximate each solution's hypervolume contribution (HVC). GHVC-Net is permutation invariant and can handle solution sets of arbitrary size, similar to the properties of GNN. Compared to HVC-Net (i.e., a machine learning model for HVC approximation), GHVC-Net achieves better accuracy with less training time. GHVC-Net is also compared with traditional approximation methods, such as line-based and point-based methods, to demonstrate its ability to identify the solution with the smallest (largest) HVC.

关键词： Training Accuracy Machine learning Approximation algorithms Graph neural networks Approximation methods Cybernetics Testing

来源：评论

学校读者我要写书评

暂无评论

Improving VTE Identification through Language Models from Radiology Reports: A Comparative Study of Mamba, Phi-3 Mini, and BERT

Improving VTE Identification through Language Models from Ra...

引用

IEEE International Conference on Bioinformatics and Biomedicine (BIBM)

作者： Jamie Deng Yusen Wu Yelena Yesha Phuong Nguyen Department of Computer Science University of Miami Miami FL USA Frost Institute for Data Science & Computing University of Miami Miami FL USA Department of Radiology University of Miami Miami FL USA

ISBN: (数字)9798350386226

ISBN: (纸本)9798350386233

Venous thromboembolism (VTE) is a critical cardio-vascular condition, encompassing deep vein thrombosis (DVT) and pulmonary embolism (PE). Accurate and timely identification of VTE is essential for effective medical care. This study builds upon our previous work, which addressed VTE detection using deep learning methods for DVT and a hybrid approach combining deep learning and rule-based classification for PE. Our earlier approaches, while effective, had two major limitations: they were complex and required expert involvement for feature engineering of the rule set. To overcome these challenges, we utilize the Mamba architecture-based classifier. This model achieves remarkable results, with a 97% accuracy and F1 score on the DVT dataset and a 98% accuracy and F1 score on the PE dataset. In contrast to the previous hybrid method on PE identification, the Mamba classifier eliminates the need for hand-engineered rules, significantly reducing model complexity while maintaining comparable performance. Additionally, we evaluated a lightweight Large Language Model (LLM), Phi-3 Mini, in detecting VTE. While this model delivers competitive results, outperforming the baseline BERT models, it proves to be computationally intensive due to its larger parameter set. Our evaluation shows that the Mamba-based model demonstrates superior performance and efficiency in VTE identification, offering an effective solution to the limitations of previous approaches.

关键词： Deep learning Training Accuracy Computational modeling Biological system modeling Computer architecture Feature extraction Complexity theory Thrombosis Biomedical imaging

来源：评论

学校读者我要写书评

暂无评论

Quantitative Error Bounds for Scaling Limits of Stochastic Iterative Algorithms

arXiv

引用

arXiv 2025年

作者： Wang, Xiaoyu Kasprzak, Mikolaj J. Negrea, Jeffrey Bourguin, Solesne Huggins, Jonathan H. Department of Mathematics & Statistics Boston University United States Department of Information Systems Data Analytics & Operations ESSEC Business School France Department of Statistics & Actuarial Science University of Waterloo Canada Faculty of Computing & Data Sciences Boston University United States

Stochastic iterative algorithms, including stochastic gradient descent (SGD) and stochastic gradient Langevin dynamics (SGLD), are widely utilized for optimization and sampling in large-scale and high-dimensional problems in machine learning, statistics, and engineering. Numerous works have bounded the parameter error in, and characterized the uncertainty of, these approximations. One common approach has been to use scaling limit analyses to relate the distribution of algorithm sample paths to a continuous-time stochastic process approximation, particularly in asymptotic setups. Focusing on the univariate setting, in this paper, we build on previous work to derive non-asymptotic functional approximation error bounds between the algorithm sample paths and the Ornstein-Uhlenbeck approximation using an infinite-dimensional version of Stein's method of exchangeable pairs. We show that this bound implies weak convergence under modest additional assumptions and leads to a bound on the error of the variance of the iterate averages of the algorithm. Furthermore, we use our main result to construct error bounds in terms of two common metrics: the Lévy-Prokhorov and bounded Wasserstein distances. Our results provide a foundation for developing similar error bounds for the multivariate setting and for more sophisticated stochastic approximation *** Codes 60F17 (Primary) 60J60, 62-08, 68T05, 62E17 (Secondary) Copyright © 2025, The Authors. All rights reserved.

关键词： Continuous time systems

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：