检索结果-内蒙古大学图书馆

10th IEEE Smart World Congress, SWC 2024

作者： Owusu, Jim Wilson Zakari, Rufai Yusuf Qin, Ke He, Tao University of Electronic Science and Technology of China School of Computer Science and Engineering China

ISBN: (纸本)9798331520861

Recent advances in graph convolutional networks (GCNs) have demonstrated their effectiveness in vision-language tasks such as visual question answering (VQA), primarily due to their ability to capture both spatial and semantic relationships. However, through a series of experiments applying GCNs to VQA tasks, we observe that the performance of GCNs is highly sensitive to various subtasks, often resulting in unstable outputs. This variability can be attributed to factors such as the size of the training and testing datasets, evaluation metrics, and the selection of hyperparameters. To ensure fair evaluations of GCN-based VQA models, it is essential to implement consistent evaluation schemes across these factors. In this study, we propose a GCN framework for VQA that leverages fine-tuned word representations from the GloVe model, specifically tailored to address reasoning-based questions. We comprehensively assess the performance of our framework using multiple evaluation metrics on both the GQA and VQA 2.0 datasets. Our experimental results demonstrate significant improvements over existing methods, showcasing the effectiveness of the proposed approach in evaluation setting. © 2024 IEEE.

关键词： Visual languages

来源：评论

学校读者我要写书评

暂无评论

Supervised Data-Dependent Kernel Low-Rank Preserving Projection for Image Recognition 5

Supervised Data-Dependent Kernel Low-Rank Preserving Project...

引用

5th International Conference on Intelligent Computing and Human-computer Interaction, ICHCI 2024

作者： Li, Linhao Zhang, Ao Jiangsu University of Science and Technology School of Computer Science and Engineering Zhenjiang China

ISBN: (纸本)9798350368284

Dimensionality reduction (DR) is a frequently used method to handle data with many dimensions. This article presents a novel approach to nonlinear dimensionality reduction, referred to as supervised data-dependent kernel low-rank preserving projection (SDKLRPP). Designed for image recognition tasks. The method enhances performance by utilizing a data-dependent kernel (DK). Unlike traditional nonlinear dimensionality reduction techniques, the SDKLRPP method described in this paper can fine-Tune the coefficients within the DK and uncover The paper elucidates the low-rank reconstructive relationship inherent in the data. It also illustrates how the manifold structure within the feature space is mirrored in the label space, grounded in the principle of smoothness. This enables label information to direct the optimization of coefficients within the DK, thereby enhancing the capacity to discern among various classes. Upon attaining the optimal low-rank reconstructive relationship, a transformation matrix is computed to maintain this relationship, which is subsequently utilized to project the data into a reduced-dimensional space. The efficacy of the proposed SDKLRPP methodology is substantiated through experiments conducted on five extensively acknowledged databases. © 2024 IEEE.

关键词： Dimensionality reduction

来源：评论

学校读者我要写书评

暂无评论

A Unimodal Valence-Arousal Driven Contrastive Learning Framework for Multimodal Multi-Label Emotion Recognition 24

A Unimodal Valence-Arousal Driven Contrastive Learning Frame...

引用

32nd ACM International Conference on Multimedia, MM 2024

作者： Zheng, Wenjie Yu, Jianfei Xia, Rui School of Computer Science and Engineering Nanjing University of Science and Technology Nanjing China

ISBN: (纸本)9798400706868

Multimodal Multi-Label Emotion Recognition (MMER) aims to identify one or more emotion categories expressed by an utterance of a speaker. Despite obtaining promising results, previous studies on MMER represent each emotion category using a one-hot vector and ignore the intrinsic relations between emotions. Moreover, existing works mainly learn the unimodal representation based on the multimodal supervision signal of a single sample, failing to explicitly capture the unique emotional state of each modality as well as its emotional correlation between samples. To overcome these issues, we propose a Unimodal Valence-Arousal driven contrastive learning framework (UniVA) for the MMER task. Specifically, we adopt the valence-arousal (VA) space to represent each emotion category and regard the emotion correlation in the VA space as priors to learn the emotion category representation. Moreover, we employ pre-trained unimodal VA models to obtain the VA scores for each modality of the training samples, and then leverage the VA scores to construct positive and negative samples, followed by applying supervised contrastive learning to learn the VA-aware unimodal representations for multi-label emotion prediction. Experimental results on two benchmark datasets MOSEI and M3ED show that the proposed UniVA framework consistently outperforms a number of existing methods for the MMER task. The source code is publicly released at https://***/NUSTM/UniVA. © 2024 ACM.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

MediServe:An IoT-Enhanced Deep Learning Framework for Personalized Medication Management for Elderly Care

引用

computers, Materials & Continua 2025年第4期83卷 935-976页

作者： Smita Kapse Ganesh Yenurkar Vincent Omollo Nyangaresi Gunjan Balpande Shravani Kale Manthan Jadhav Sahil Lawankar Vikrant Jaunjale Department of Computer Technology Yeshwantrao Chavan College of EngineeringWanadongariNagpur441110MaharashtraIndia Computer Science&Engineering Jaramogi Oginga Odinga University of Science&TechnologyBondo40601Kenya Department of Applied Electronics Saveetha School of EngineeringSIMATSChennai602105TamilnaduIndia

In today’s fast-paced world,many elderly individuals struggle to adhere to their medication schedules,especially those with memory-related conditions like Alzheimer’s disease,leading to serious health risks,hospital-izations,and increased healthcare *** reminder systems often fail due to a lack of personalization and real-time *** address this critical challenge,we introduce MediServe,an advanced IoT-enabled medication management system that seamlessly integrates deep learning techniques to provide a personalized,secure,and adaptive *** features a smart medication box equipped with biometric authentication,such as fingerprint recognition,ensuring authorized access to prescribed medication while preventing misuse.A user-friendly mobile application complements the system,offering real-time notifications,adherence tracking,and emergency alerts for caregivers and healthcare *** system employs predictive deep learning models,achieving an impressive classification accuracy of 98%,to analyze user behavior,detect anomalies in medication adherence,and optimize scheduling based on an individual’s habits and health ***,MediServe enhances accessibility by employing natural language processing(NLP)models for voice-activated interactions and text-to-speech capabilities,making it especially beneficial for visually impaired users and those with cognitive ***-based data analytics and wireless connectivity facilitate remote monitoring,ensuring that caregivers receive instant alerts in case of missed doses or medication ***,machine learning-based clustering and anomaly detection refine medication reminders by adapting to users’changing health *** combining IoT,deep learning,and advanced security protocols,MediServe delivers a comprehensive,intelligent,and inclusive solution for medication *** innovative approach not only improves the quality of life for elderly

关键词： MediServe medication health risks smart medication box

来源：评论

学校读者我要写书评

暂无评论

Self-Supervised Learning Recommendation with Enhanced Long-Tail Nodes 5

Self-Supervised Learning Recommendation with Enhanced Long-T...

引用

5th International Conference on Artificial Intelligence and computer engineering, ICAICE 2024

作者： Zhao, Zhiyu Wang, Yongli Liu, Dongmei School of Computer Science and Engineering Nanjing University of Science and Technology Nanjing China

ISBN: (纸本)9798331528911

Achieving a balance between recommending popular items and long-tail items has consistently posed a challenge in the field of recommendation systems. Traditional recommendation algorithms often exhibit a bias towards recommending popular items, resulting in limited exposure for long-tail items. To address this issue, our study introduces an innovative recommendation algorithm based on graph contrastive learning. This algorithm incorporates a dual-view learning mechanism with the objective of increasing the exposure of long-tail items while ensuring the accuracy of recommendations. Two distinct views are generated: one captures in-depth information of long-tail nodes through random walk strategies, while the other utilizes Graph Attention Networks (GAT) to acquire node importance and contextual information. By means of contrastive learning with these two views, our model attains comprehensive node representations. The study strives to achieve a balanced learning process between long-tail and high-frequency nodes. Experimental results conducted on multiple recommendation system datasets substantiate that our approach significantly enhances the exposure of long-tail items compared to existing recommendation algorithms, while maintaining high accuracy and recall in recommendations. This research attests that the contrastive learning framework provides an effective strategy for addressing the long-tail recommendation problem. © 2024 IEEE.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Learning in sinusoidal spaces with physics-informed neural networks

IEEE Transactions on Artificial Intelligence

引用

IEEE Transactions on Artificial Intelligence 2024年第3期5卷 985-1000页

作者： Wong, Jian Cheng Ooi, Chin Chun Gupta, Abhishek Ong, Yew Soon Institute of High Performance Computing Agency for Science Technology and Research Singapore138632 Singapore School of Computer Science and Engineering Nanyang Technological University Singapore639798 Singapore Singapore Institute of Manufacturing Technology Agency for Science Technology and Research Singapore138632 Singapore Agency for Science Technology and Research Singapore138632 Singapore Data Science and Artificial Intelligence Research Centre School of Computer Science and Engineering Nanyang Technological University Singapore138632 Singapore

A physics-informed neural network (PINN) uses physics-Augmented loss functions, e.g., incorporating the residual term from governing partial differential equations (PDEs), to ensure its output is consistent with fundamental physics laws. However, it turns out to be difficult to train an accurate PINN model for many problems in practice. In this article, we present a novel perspective of the merits of learning in sinusoidal spaces with PINNs. By analyzing behavior at model initialization, we first show that a PINN of increasing expressiveness induces an initial bias around flat output functions. Notably, this initial solution can be very close to satisfying many physics PDEs, i.e., falling into a localminimum of the PINN loss that onlyminimizes PDE residuals, while still being far from the true solution that jointly minimizes PDE residuals and the initial and/or boundary conditions. It is difficult for gradient descent optimization to escape from such a local minimum trap, often causing the training to stall. We then prove that the sinusoidalmapping of inputs-in an architecture we label as sf-PINN-is effective to increase input gradient variability, thus avoiding being trapped in such deceptive local minimum. The level of variability can be effectively modulated to match high-frequency patterns in the problem at hand. A key facet of this article is the comprehensive empirical study that demonstrates the efficacy of learning in sinusoidal spaces with PINNs for a wide range of forward and inversemodeling problems spanning multiple physics domains. Impact Statement-Falling under the emerging field of physicsinformed machine learning, PINN models have tremendous potential as a unifying AI framework for assimilating physics theory and measurement data. However, they remain infeasible for broad science and engineering applications due to computational cost and training challenges, especially for more complex problems. Instead of focusing on empirical demonstration of appli

关键词： Gradient methods

来源：评论

学校读者我要写书评

暂无评论

Inverse design of winding tuple for non-Hermitian topological edge modes

引用

Physical Review B 2025年第4期111卷 L041406-L041406页

作者： Zihe Yang Kunling Zhou Bowen Zeng Yong Hu School of Physics Huazhong University of Science and Technology Wuhan 430074 People's Republic of China Hunan Provincial Key Laboratory of Flexible Electronic Materials Genome Engineering School of Physics and Electronic Sciences Changsha University of Science and Technology Changsha 410114 People's Republic of China

The interplay between topological localization and non-Hermiticity localization in non-Hermitian crystal systems results in a diversity of shapes of topological edge modes (EMs), offering opportunities to manipulate these modes for potential topological applications. The characterization of the domain of EMs and the engineering of these EMs require detailed information about their wave function, which conventional calculation of topological invariants cannot provide. In this Letter, by recognizing EMs as specified solutions of an eigenequation, we derive their wave functions in an extended non-Hermitian Su-Schrieffer-Heeger model. We then inversely construct a winding tuple {wGBZ,wBZ} that characterizes the existence of EMs and their spatial distribution. Moreover, we define a spectral winding number equivalent to wBZ, which is determined by the product of energies of different bands. The inverse design of topological invariants allows us to categorize the localized nature of EMs even in systems lacking sublattice symmetry, which can facilitate the manipulation and utilization of EMs in the development of different quantum materials and devices.

关键词： Wave functions

来源：评论

学校读者我要写书评

暂无评论

Q2Cypher: Converting Natural Language Questions to Cypher with Fine-Tuned Large Language Models 5

Q2Cypher: Converting Natural Language Questions to Cypher wi...

引用

5th International Conference on Artificial Intelligence and computer engineering, ICAICE 2024

作者： Zou, Yunqi Wang, Yongli Liu, Dongmei School of Computer Science and Engineering Nanjing University of Science and Technology Nanjing China

ISBN: (纸本)9798331528911

Knowledge Base Question Answering (KBQA) intends to obtain credible answers to natural language questions based on knowledge bases. Knowledge graphs, also known as graph databases, explicitly present a large amount of factual knowledge and the associations among them. Neo4j, one of the most popular graph databases, uses the Cypher Query Language (CQL) for efficient querying. Due to its complexity, converting natural language questions into CQL, i.e., Text-to-CQL, is in demand, but the performance is undesirable for existing models. To address the issue, we propose Q2Cypher, a Text-to-CQL framework that is able to generate corresponding CQL based on natural language questions by fine-tuning open-source large language models (LLMs). Q2Cypher employs an unsupervised joint retrieval method to further align entity and relation names in the CQL, thus improving generation from the models more directly. We conduct experiments on open-source LLMs such as ChatGLM2, Baichuan2 and Llama-2, using different fine-tuning methods such as LoRA, Freeze and QLoRA for comparison. The experimental results demonstrate that Q2Cypher is effective and achieves better performance than existing methods on the SpCQL dataset. © 2024 IEEE.

关键词： Knowledge graph

来源：评论

学校读者我要写书评

暂无评论

A Novel Knowledge Enhanced Large Language Model Augmented Framework for Medical Question Answering

A Novel Knowledge Enhanced Large Language Model Augmented Fr...

引用

2024 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2024

作者： Zou, Haochen Wang, Yongli Nanjing University of Science and Technology School of Computer Science and Engineering Nanjing China

ISBN: (纸本)9798350386226

Leveraging domain-specific knowledge from pre-trained large language models and knowledge graphs for reasoning in the medical question answering task has emerged as a prominent research field. However, the accuracy of the inference results is restricted by multiple factors, including the quality of analyzed topic entities, the selected inference path in the knowledge graph, and the absence of mutual updating for embedding representations from large language models and knowledge graphs. In this paper, we propose a novel medical question answering framework based on the domain-specific large language model, aiming to enhance the quality of topic entities by implementing the retrieval augmentation technique. Inspired by the concept of chain-of-thought reasoning, we introduce a joint reasoning approach based on analyzed topic entities to facilitate the generation of accurate inference paths. Additionally, we design a unified embedding mechanism that combines representations from both the large language model and graph neural networks, incorporating a pooling operation for predicting answers to input questions. To the best of our knowledge, this work signifies the pioneering efforts in implementing the retrieval augmentation technique and the joint reasoning approach within the context of the medical question answering task. Experimental results on public benchmark datasets demonstrate that the introduced method outperforms state-of-the-art baseline approaches. © 2024 IEEE.

关键词： Knowledge graph

来源：评论

学校读者我要写书评

暂无评论

Leveraging the Potential of Large Language Models

引用

Informatica (Slovenia) 2024年第8期48卷 1-16页

作者： Prasad, Shreya Gupta, Himank Ghosh, Arup Department of Software Systems School of Computer Science and Engineering Vellore Institute of Technology Tamil Nadu Vellore India Department of Computer Science and Information Technology Graceland University LamoniIA United States

This study focuses on enhancing Natural Language Processing (NLP) in generative AI chatbots through the utilization of advanced pre-trained models. We assessed five distinct Large Language Models (LLMs): TRANSFORMER MODEL, FALCON 7B, LAMINI-FLAN-T5-783M, LLAMA-2-7B, and LLAMA-2-13B to identify the most effective one. Our findings revealed that the LLAMA Model excels in comprehending user queries and delivering precise responses during conversations. The article elucidates the methodology employed to evaluate and select various models for our chatbot. Through rigorous testing, we determined that the LLAMA-2-13B model exhibits enhanced response time and accuracy. Additionally, we employed tools such as Facebook Artificial Intelligence Similarity Search (FAISS) and experimented with user interfaces like Streamlit and Chainlit to enhance the chatbot's user-friendliness. The research underscores the significance of selecting the appropriate model for crafting efficient chatbots. Ultimately, the LLAMA-13B model emerged as the standout performer, showcasing superior performance. Benchmark assessments, including HellaSwag and WinoGrande, which gauge common sense reasoning, were employed to evaluate our chatbot's capabilities. The study concludes that LLAMA-based models hold significant promise for the development of innovative and user-friendly chatbots in the future. Povzetek: Študija se osredotoča na izboljšanje obdelave naravnega jezika (NLP) v generativnih AI klepetalnicah z uporabo naprednih vnaprej usposobljenih modelov. Analizirali so pet modelov: TRANSFORMER MODEL, FALCON 7B, LAMINI-FLAN-T5-783M, LLAMA-2-7B in LLAMA-2-13B. Model LLAMA-13B se je izkazal kot najboljši z vrhunskimi rezultati na testih HellaSwag in WinoGrande. Raziskava uporablja orodja, kot je FAISS, in uporabniške vmesnike, kot sta Streamlit in Chainlit za izboljšanje uporabniške izkušnje. © 2024 Slovene Society Informatika. 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：