检索结果-内蒙古大学图书馆

2024 Conference on Empirical Methods in Natural Language Processing, EMNLP 2024

作者： Wei, Zhaohui Liao, Lizi Du, Xiaoyu Xiang, Xinguang School of Computer Science and Engineering Nanjing University of Science and Technology China Singapore Management University Singapore

ISBN: (纸本)9798891761681

In the realm of dialogue-to-image retrieval, the primary challenge is to fetch images from a pre-compiled database that accurately reflect the intent embedded within the dialogue *** methods often overemphasize inter-modal alignment, neglecting the nuanced nature of conversational *** histories are frequently cluttered with redundant information and often lack direct image descriptions, leading to a substantial disconnect between conversational content and visual *** study introduces VCU, a novel framework designed to enhance the comprehension of dialogue history and improve cross-modal matching for image *** leverages large language models (LLMs) to perform a two-step extraction *** generates precise image-related descriptions from dialogues, while also enhancing visual representation by utilizing object-list texts associated with ***, auxiliary query collections are constructed to balance the matching process, thereby reducing bias in similarity *** results demonstrate that VCU significantly outperforms baseline methods in dialogue-to-image retrieval tasks, highlighting its potential for practical application and effectiveness in bridging the gap between dialogue context and visual content. © 2024 Association for Computational Linguistics.

关键词： Computational linguistics

来源：评论

学校读者我要写书评

暂无评论

Dual-Branch Differentiated Similarity Network for Semi-supervised Medical Image Segmentation 19th

Dual-Branch Differentiated Similarity Network for Semi-supe...

引用

19th EAI International Conference on Heterogeneous Networking for Quality, Reliability, Security and Robustness, QShine 2023

作者： Yang, Weixian Lin, Jing Cai, Wentian Gao, Ying School of Computer Science and Engineering South China University of Technology Guangzhou510006 China

ISBN: (纸本)9783031651229

Consistency learning has been successfully applied in semi-supervised medical image segmentation, as it enforces consistency in model prediction across different perturbations or transformations of the same input data. The accuracy of consistent learning is challenged by predictive diversity and model training stability, which are often neglected by existing researches. Meanwhile, the potential inter-class similarities between labeled and unlabeled training data are not effectively mined. To address these issues, we propose a semi-supervised framework based on dual decoders, namely Dual-Branch Differentiated Similarity Network (DBDSNet). First, dual-branch network and cross pseudo supervision can enable model to learn more invariant and meaningful representations from the data. Second, we proposed a Differentiated Similarity Loss (DSL) to encourage dual branches to focus on capturing the semantic information of the data, rather than relying on the noisy pseudo label. Last, we propose a Inter-Class Consistency Module (ICCM) to mine the inter-class similarity between labeled data and unlabelled data. Extensive experiments conducted on two public medical image datasets demonstrate that our method reaches state-of-the-art performance compared with former methods. © ICST Institute for computer sciences, Social Informatics and Telecommunications engineering 2024.

关键词： Self-supervised learning

来源：评论

学校读者我要写书评

暂无评论

Channel modeling for MI-based wireless underground sensor networks with conductive objects

引用

science China(Information sciences) 2024年第7期67卷 333-334页

作者： Shuhan DENG Guanghua LIU Ziwei CHEN Huaijin ZHANG Tao JIANG Research Center of 6G Mobile Communications School of Cyber Science and EngineeringHuazhong University of Science and Technology School of Electronic Information and Communications Huazhong University of Science and Technology

Wireless underground sensor networks (WUSNs) are an emerging research field with promising applications. However, WUSNs links suffer from significant attenuation as the propagation of the signal needs to pass through soil, rock,and water. Magnetic induction (MI)-based transmission is proposed to overcome these issues as MI signal can penetrate most substances similarly without experiencing high path loss, which ensures the robustness of WUSNs links [1].

关键词： Wireless sensor networks

来源：评论

学校读者我要写书评

暂无评论

An Innovative Coupled Common-Node Discrete Element Method-Smoothed Particle Hydrodynamics Model Developed with LS-DYNA and Its Applications

引用

China Ocean engineering 2024年第3期38卷 467-482页

作者： SHEN Zhong-xiang WANG Wen-qing XU Cheng-yue LUO Jia-xin LIU Ren-wei School of Naval Architecture and Ocean Engineering Jiangsu University of Science and TechnologyZhenjiang 212100China College of Harbor Coastal and Offshore EngineeringHohai UniversityNanjing 210098China

In this study,a common-node DEM-SPH coupling model based on the shared node method is proposed,and a fluid–structure coupling method using the common-node discrete element method-smoothed particle hydrodynamics(DS-SPH)method is developed using LS-DYNA *** DEM and SPH are established on the same node to create common-node DEM-SPH particles,allowing for fluid–structure *** simulations of various scenarios,including water entry of a rigid sphere,dam-break propagation over wet beds,impact on an ice plate floating on water and ice accumulation on offshore structures,are *** interaction between DS particles and SPH fluid and the crack generation mechanism and expansion characteristics of the ice plate under the interaction of structure and fluid are also *** results are compared with available data to verify the proposed coupling ***,the simulation results demonstrated that controlling the cutoff pressure of internal SPH particles could effectively control particle splashing during ice crushing failure.

关键词： common-node DEM-SPH fluid-structure interaction discrete element method smoothed particle hydrodynamics

来源：评论

学校读者我要写书评

暂无评论

PPO-Based Mobile Robot Path Planning with Dense Bootstrap Reward in Partially Observable Environments 21

PPO-Based Mobile Robot Path Planning with Dense Bootstrap Re...

引用

21st International computer Conference on Wavelet Active Media technology and Information Processing, ICCWAMTIP 2024

作者： Shujie, Zhou School of Computer Science and Engineering School of Cyber Security University of Electronic Science and Technology of China Chengdu611731 China

ISBN: (纸本)9798331519254

Path planning in large, complex environments poses significant challenges, as a single robot must efficiently navigate to its target while avoiding obstacles. Traditional algorithms often struggle to balance computational efficiency and planning accuracy. Reinforcement learning methods, while promising, frequently suffer from convergence issues due to difficult-to-tune hyperparameters and sparse reward signals. This paper introduces a Proximal Policy Optimization (PPO)-based path planning approach with a dense bootstrap reward mechanism. The proposed method enhances learning efficiency and ensures smooth navigation by incorporating a structured reward design, significantly improving the agent's path-planning performance in complex scenarios. We conducted simulation experiments in a classical warehouse environment to demonstrate the effectiveness and superiority of our proposed method. Compared to traditional rewards, our method significantly improves the pathfinding efficiency of deep reinforcement learning algorithms. © 2024 IEEE.

关键词： Deep reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Prediction and Management of Building Energy Consumption Based on Building Environment Simulation Design Platform DeST and Meteorological Data Analysis Algorithm

引用

Strategic Planning for Energy and the Environment 2024年第2期43卷 357-380页

作者： Bai, Chaoqin Liu, Junrui School of Civil Engineering and Architecture Henan University of Science and Technology Luoyang471000 China

Currently, the carbon emissions of building energy consumption account for a significant portion of all carbon emissions. How to reduce carbon emissions to achieve carbon neutrality is an important current research direction. Therefore this research builds a predictive algorithm model for analyzing energy consumption data of meteorological buildings using DeST platform for energy saving and emission reduction to achieve carbon neutrality. The new model uses Internet of Things and cloud platform technology to build a simulation building platform, and uses the support vector machine algorithm in the analysis algorithm to vectorize building energy consumption data, which can achieve normalization processing of building energy consumption and meteorological data. By processing building energy consumption data, prediction of building energy consumption at the next moment can be achieved. The experimental results show that the precision and accuracy of the new algorithm are higher than genetic algorithm 1 and 0.15 respectively, and 0.6 and 0.07 higher than clustering analysis algorithm respectively. Therefore, applying this algorithm model to building energy consumption prediction can significantly improve the accuracy and precision of the algorithm. © 2024 River Publishers.

关键词： Carbon sequestration

来源：评论

学校读者我要写书评

暂无评论

Sentiment Analysis Unleashed: Leveraging VADER for Evaluation of Public Opinion 11

Sentiment Analysis Unleashed: Leveraging VADER for Evaluatio...

引用

11th International Conference on Reliability, Infocom Technologies and Optimization, ICRITO 2024

作者： Arora, Suhawni Malhotra, Ashmit Kumar, Amit Singh, Ninni University Department of Computer Science Engineering & Technology Dehradun India Cmr Institute of Technology Department of Computer Science Engineering & Technology Telangana Hyderabad India

ISBN: (纸本)9798350350357

Digital reviews provide real-world feedback on products and services in an era of online commerce and access to information. Providing feedback fosters trust and credibility among potential customers, enabling them to make informed purchasing decisions. In review classification, sentiment analysis is a key component that assesses the emotional tone of the reviews. Businesses can prioritize responses, manage their reputation, and extract actionable insights by categorizing reviews as positive, negative, or neutral. The study also includes neutral feedback, which includes suggestions for incremental improvement, which assists in continuous product refinement. In the proposed work, more than 5,68,455 reviews are divided into positive negative, and neutral sentiments by sentiment analysis. The results demonstrate VADER's robustness and versatility, showing its capacity to accurately gauge public sentiment across different contexts and types of language, including slang, emojis, and colloquial expressions. We also compare VADER's performance with other sentiment analysis algorithms, underscoring its advantages in handling informal online communication. © 2024 IEEE.

关键词： Sentiment analysis

来源：评论

学校读者我要写书评

暂无评论

A Review on Meta-Learning: How Artificial Intelligence and Machine Learning Can Learn to Adapt Quickly 3

A Review on Meta-Learning: How Artificial Intelligence and M...

引用

3rd International Conference on Electronics and Renewable Systems, ICEARS 2025

作者： Khadse, Shrikant Gourshettiwar, Palash Pawar, Adesh Faculty of Engineering and Technology Wardha442001 India Faculty of Engineering and Technology Department of Computer Science and Medical Engineering Wardha442001 India Department of Computer Science and Medical Engineering Maharashtra Wardha442001 India

ISBN: (纸本)9798331509675

Meta-learning aims to create Artificial Intelligence (AI) systems that can adapt to new tasks and improve their performance over time without extensive retraining. The advent of meta-learning paradigms has fundamentally changed how Artificial Intelligence (AI) and Machine Learning (ML) are handled, constantly reshaping these fields. Research study reviews the broad field of meta-learning, highlighting its critical function in bridging the AI-ML divide to improve adaptability. The study starts with a review of past developments, following the development of meta-learning algorithms and their underpinning theories. The historical view lays the groundwork for understanding the course of this dynamic regulation, spanning from early efforts in probabilistic program induction to modern discoveries in model-agnostic meta-learning. The study explores multiple meta-learning applications in AI and ML, revealing how they affect financial forecasting, computer vision, natural language processing, autonomous cars, health informatics, robotics, and personalized recommendation systems. When seen from the perspective of adaptive intelligence, meta-learning may be effectively applied to conditions with limited data, hyperparameter optimization, few-shot learning, and quick adaption in reinforcement learning settings. © 2025 IEEE.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Finger vein recognition based on efficient channel attention and GhostNet

引用

International Journal of Wireless and Mobile Computing 2024年第2期26卷 157-167页

作者： Ke, Yintao Zheng, Hui Jie, Jing Hou, Beiping Chen, Yuchuan School of Automation and Electrical Engineering Zhejiang University of Science and Technology Zhejiang Hangzhou China

The deep convolutional network (DCN) suffers from drawbacks such as high computational complexity and slow speed. To address these issues and facilitate the deployment of DCNs in embedded devices, we propose a finger vein recognition method based on lightweight Efficient Channel Attention (ECA) mechanism and GhostNet. By combining the ECA mechanism with the GhostNet’s G-bneck, we create a new module called ECAGhostNet. Additionally, we establish a more realistic FV-UST dataset for finger vein door locks which includes images with various challenges like rotation, stains, skin damage, hand sweat, temperature variations, and illumination differences. Experimental results demonstrate that ECAGhostNet outperforms GhostNet on the public FV-USM dataset, improving accuracy by 0.82% with minimal parameter increase (1.9 M). On the self-built FV-UST dataset, ECAGhostNet achieves a 0.63% accuracy improvement over GhostNet. Furthermore, we validate the effectiveness of our proposed model on the Jetson Nano device, confirming its suitability for real-world embedded applications. © 2024 Inderscience Enterprises Ltd.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

A Hybrid Framework for Data Loss Prevention and Detection

A Hybrid Framework for Data Loss Prevention and Detection

引用

2024 IEEE International Conference on Signal Processing and Advance Research in Computing, SPARC 2024

作者： Srivastava, Atul Srivastava, Priyam Pillai, Anuradha Sharma, Vijay Shankar AUUP Amity School of Engineering and Technology Lucknow India Symbiosis Institute of Technology Department of Computer Science and Engineering Pune India Manipal University Department of Computer and Communication Engineering Jaipur India

ISBN: (纸本)9798350385199

Modern enterprises face a serious threat from data loss, or the unintentional or undesired disclosure of data. Modern Data Loss Protection (DLP) systems either discover anomalies in regular behavior (anomaly-based) or use signatures of known attack patterns. Though signature-based solutions are effective in accurately identifying known attacks and may thus be used to block them, they are unable to handle unexpected attacks or attackers who choose atypical attack vectors (such as those that are only known to insiders) to execute their attack. However, because they frequently produce a high false positive rate, anomaly-based solutions are not as useful for identifying suspicious activity. In contrast, they can identify undiscovered attacks. Our research presents a hybrid DLP system that integrates anomaly-based and signature-based approaches to facilitate both prevention and detection. Using an anomaly-based engine, the framework automatically builds a model of typical user behavior, enabling it to detect unusual transactions carried out by insiders. This is usually where anomaly-based solutions end. To prevent unwanted transactions from causing harm, our framework goes one step further by using an operator's feedback on alarms to automatically create and update attack signatures. We manage to improve time to detect a data breach was reduced by 67%, improving response efficiency. © 2024 IEEE.

关键词： Information leakage

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：