检索结果-内蒙古大学图书馆

Connecting Cross-Modal Representations for Compact and Robust Multimodal Sentiment Analysis with Sentiment Word Substitution Error

引用

IEEE Transactions on Affective Computing 2024年

作者： Sun, Qiyuan Zuo, Haolin Liu, Rui Li, Haizhou The Department of Computer Science Inner Mongolia University Hohhot010021 China Shenzhen Research Institute of Big Data School of Data Science The Chinese University of Hong Kong CUHKShenzhen Shenzhen518172 China University of Bremen Faculty3 Computer Science / Mathematics Enrique-Schmidt-Str. 5 Cartesium Bremen28359 Germany

Multimodal Sentiment Analysis (MSA) seeks to fuse textual, acoustic, and visual information to predict a speaker’s sentiment states effectively. However, in real-world scenarios, the text modality received by MSA systems is often obtained through automatic speech recognition (ASR) models. Unfortunately, ASR may erroneously recognize sentiment words as phonetically similar neutral alternatives, leading to sentiment degradation in text and impacting MSA accuracy. Recent attempts aim to first identify the sentiment word substitution (SWS) error in ASR results and then refine the corrupted word embeddings using multimodal information for final multimodal fusion. However, such a method includes a burdensome and ambiguous detection operation and ignores the inherent correlations and heterogeneity among different modalities. To address these issues, we propose a more compact system, termed ARF-MSA consisting of three key components to achieving robust MSA with SWS errors: 1) Alignment: we establish connections between the "text-acoustic’ and "text-visual" representations to effectively map the "text-acoustic-visual" data into a unified sentiment space by leveraging their multimodal correlation knowledge;2) Refinement: we perform fine-grained comparisons between the text modality and the other two modalities in the unified sentiment space, enabling refinement of the sentiment expression within the text modality more concisely;3) Fusion: Finally, we hierarchically fuse the dominant and non-dominant representation from three heterogeneity modalities to obtain the multimodal feature for MSA. We conduct extensive experiments on the real-world datasets and the results demonstrate the effectiveness of our model. © 2010-2012 IEEE.

关键词： Alignment

来源：评论

学校读者我要写书评

暂无评论

Personalized Care for Multiple Sclerosis with IoT Wearables and Deep Learning

Personalized Care for Multiple Sclerosis with IoT Wearables ...

引用

2024 Asian Conference on Intelligent Technologies, ACOIT 2024

作者： Vetriselvi, D. Sampathrajan, Rajeshkumar Deepa, A. Sujatha, V. Deepa, P. Pandi Murugan, S. Department of Computing Technologies School of Computing Faculty of Engineering and Technology SRM Institute of Science and Technology Kattankulathur Tamil Nadu Chennai India McKinsey & Company Fort WorthTX United States Department of Computer Science and Engineering Sathyabama Institute of Science and Technology Tamil Nadu Chennai India Department of Electronics and Communication Engineering Shree Sathyam College of Engineering and Technology Tamil Nadu Salem India Department of Artificial Intelligence and Data Science Agni College of Technology Tamil Nadu Chennai India Department of Computer Science and Engineering Saveetha School of Engineering Saveetha Institute of Medical and Technical Sciences Saveetha University Tamil Nadu Chennai India

ISBN: (纸本)9798350374933

This research work presents a new method for improving individualized treatment for patients with Multiple Sclerosis (MS) by combining Internet of Things (IoT) wearables with deep learning (DL) methods, namely Long Short-Term Memory (LSTM) networks. MS is a long-term autoimmune disorder that affects the central nervous system. It is marked by unexpected episodes of worsening symptoms and periods of improvement, highlighting the need for tailored therapy. The widespread use of IoT devices presents unique possibilities for the ongoing monitoring of many physiological and activity measurements important to the development and symptoms of MS. Using LSTM networks, which are very proficient in analyzing temporal sequences, create a prediction model that may anticipate exacerbations of MS symptoms and customize therapies appropriately. Through the analysis of longitudinal data obtained from IoT devices, our model acquires knowledge about unique patterns of illness development and treatment response. This knowledge allows us to create customized care plans for individuals. These paths include proactive activities such as adjusting medication, making lifestyle changes, and seeking early medical advice. The goal is to reduce the worsening of symptoms, improve the quality of life, and optimize treatment results for individuals with MS. The proposed method shows potential for transforming the treatment of MS by offering proactive and patient-centered therapy adaptable to individual requirements and the dynamics of the illness. © 2024 IEEE.

关键词： Patient treatment

来源：评论

学校读者我要写书评

暂无评论

YOLOv5 Algorithm Optimization: Precise Target Detection Amidst Background Interference 4

YOLOv5 Algorithm Optimization: Precise Target Detection Amid...

引用

4th International Conference on Consumer Electronics and computer Engineering, ICCECE 2024

作者： Xiong, Bingbing Shi, Wenxi Fu, Yanyun Zhu, Zhansheng Wang, Deyong Jia, Bin School of Computer Science and Technology Xinjiang University Urumqi China Xinjiang Lianhaichuangzhi Information Technology Co. Ltd. Key Laboratory of Big Data of Xinjiang Social Security Risk Beijing China Beijing Academy of Science and Technology Beijing China Xinjiang Lianhaichuangzhi Information Technology Co. Ltd. Key Laboratory of Big Data of Xinjiang Social Security Risk Urumqi China

ISBN: (纸本)9798350384062

This paper solves the challenge that the target is submerged in the background in the border security scenario, and proposes the enhancement of YOLOv5 algorithm. EfficientViT replaces the DarkNet backbone, leveraging Transformer models for improved feature extraction. Modifications to the Spatial Pyramid Pooling-Fast (SPPF) module enhance the algorithm's robustness to scale variations. Introducing SCYLLA-IoU (SIoU) in the loss function improves model convergence speed and accuracy. Validated on a custom border security dataset, the enhancements show an improved average precision mean (mAP) of 48.3%, a 2.2 percentage point increase over the original YOLOv5s model. © 2024 IEEE.

关键词： Border security Target Detection Transformer YOLOv5

来源：评论

学校读者我要写书评

暂无评论

Procedural Fairness and Its Relationship with Distributive Fairness in Machine Learning

arXiv

引用

arXiv 2025年

作者： Wang, Ziming Huang, Changwu Tang, Ke Yao, Xin Guangdong Provincial Key Laboratory of Brain-inspired Intelligent Computation Department of Computer Science Engineering Southern University of Science and Technology Shenzhen518055 China Department of Computer Science and Engineering Southern University of Science and Technology Shenzhen518055 China School of Data Science Lingnan University Hong Kong

Fairness in machine learning (ML) has garnered significant attention in recent years. While existing research has predominantly focused on the distributive fairness of ML models, there has been limited exploration of procedural fairness. This paper proposes a novel method to achieve procedural fairness during the model training phase. The effectiveness of the proposed method is validated through experiments conducted on one synthetic and six real-world datasets. Additionally, this work studies the relationship between procedural fairness and distributive fairness in ML models. On one hand, the impact of dataset bias and the procedural fairness of ML model on its distributive fairness is examined. The results highlight a significant influence of both dataset bias and procedural fairness on distributive fairness. On the other hand, the distinctions between optimizing procedural and distributive fairness metrics are analyzed. Experimental results demonstrate that optimizing procedural fairness metrics mitigates biases introduced or amplified by the decision-making process, thereby ensuring fairness in the decision-making process itself, as well as improving distributive fairness. In contrast, optimizing distributive fairness metrics encourages the ML model’s decision-making process to favor disadvantaged groups, counterbalancing the Copyright © 2025, The Authors. All rights reserved.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Deep Learning Approaches for Solving Challenges in computer Vision Applications and Exploring Evolving Patterns 7

Deep Learning Approaches for Solving Challenges in Computer ...

引用

7th International Conference on Contemporary Computing and Informatics, IC3I 2024

作者： Gejji, Sujay Supreeth, B.R. Khan, Pathan Firoze Venkat, S. Pradeepa, K. Kaliappan, S. S. G. Balekundri Institute of Technology Department of Electronics & Communication Engineering Karnataka Belagavi India Oxford Degree and Pg College Department of Computer Science Sir. M. Visveshwaraiah Layout Bangalore India Madanapalle institute of Technology & Science Department of Computer Science & Technology Andhra Pradesh Madanapalle India S. A. Engineering College Department of Electronics and communication Engineering Tamil Nadu Chennai India Kpr College of Arts Science and Research Department of Artificial Intelligence and Data Science Tamil Nadu Coimbatore India Lovely Professional University Division of Research and Development Punjab Phagwara India

ISBN: (纸本)9798350350067

Deep learning has brought substantial advancements to computer vision in recent years by providing answers to challenging visual interpretation problems. It investigates several DL methodologies, such as self-supervised learning strategies, hybrid CNN-RNN models, enhanced CNNs, generative adversarial networks (GANs), and vision transformers (ViTs). Novel CNN architectures are designed, GANs are used for image synthesis, hybrid models are used to integrate spatial and temporal information, self-attention is used in ViTs, and self-supervised learning is used for feature extraction. The results of the study show significant improvements: the accuracy of Enhanced CNNs was 81.6%, the accuracy of GANs was 14.8%, the accuracy of Hybrid CNN-RNN models was 86.2%, the accuracy of ViTs was 79.2%, and the accuracy of transfer for self-supervised learning techniques was 76.8%. These findings demonstrate the improved effectiveness and performance of the suggested techniques, highlighting their potential to resolve challenging problems in feature extraction, video analysis, object recognition, picture enhancement, and image classification. It concludes that advances in computer vision will be facilitated by the ongoing development of DL methods. These advancements will lead to more precise and adaptable applications in a range of fields, including autonomous driving, medical imaging, surveillance, and augmented reality. © 2024 IEEE.

关键词： Generative adversarial networks

来源：评论

学校读者我要写书评

暂无评论

Stabilize the latent space for image autoregressive modeling: a unified perspective 24

Stabilize the latent space for image autoregressive modeling...

引用

Proceedings of the 38th International Conference on Neural Information Processing Systems

作者： Yongxin Zhu Bocheng Li Hang Zhang Xin Li Linli Xu Lidong Bing School of Data Science University of Science and Technology of China and State Key Laboratory of Cognitive Intelligence School of Computer Science and Technology University of Science and Technology of China and State Key Laboratory of Cognitive Intelligence Zhejiang University

ISBN: (纸本)9798331314385

Latent-based image generative models, such as Latent Diffusion Models (LDMs) and Mask Image Models (MIMs), have achieved notable success in image generation tasks. These models typically leverage reconstructive autoencoders like VQGAN or VAE to encode pixels into a more compact latent space and learn the data distribution in the latent space instead of directly from pixels. However, this practice raises a pertinent question: Is it truly the optimal choice? In response, we begin with an intriguing observation: despite sharing the same latent space, autoregressive models significantly lag behind LDMs and MIMs in image generation. This finding contrasts sharply with the field of NLP, where the autoregressive model GPT has established a commanding presence. To address this discrepancy, we introduce a unified perspective on the relationship between latent space and generative models, emphasizing the stability of latent space in image generative modeling. Furthermore, we propose a simple but effective discrete image tokenizer to stabilize the latent space for image generative modeling by applying K-Means on the latent features of self-supervised learning models. Experimental results show that image autoregressive modeling with our tokenizer (DiGIT) benefits both image understanding and image generation with the next token prediction principle, which is inherently straightforward for GPT models but challenging for other generative models. Remarkably, for the first time, a GPT-style autoregressive model for images outperforms LDMs, which also exhibits substantial improvement akin to GPT when scaling up model size. Our findings underscore the potential of an optimized latent space and the integration of discrete tokenization in advancing the capabilities of image generative models. The code is available at https://***/DAMO-NLP-SG/DiGIT.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Non-Convex Tensor Recovery from Local Measurements 39

Non-Convex Tensor Recovery from Local Measurements

引用

39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025

作者： Wu, Tongle Sun, Ying Fan, Jicong School of Electrical Engineering and Computer Science Pennsylvania State University United States School of Data Science The Chinese University of Hong Kong Shenzhen China

ISBN: (纸本)157735897X

Motivated by the settings where sensing the entire tensor is infeasible, this paper proposes a novel tensor compressed sensing model, where measurements are only obtained from sensing each lateral slice via mutually independent matrices. Leveraging the low tubal rank structure, we reparameterize the unknown tensor X★ using two compact tensor factors and formulate the recovery problem as a nonconvex minimization problem. To solve the problem, we first propose an alternating minimization algorithm, termed Alt-PGD-Min, that iteratively optimizes the two factors using a projected gradient descent and an exact minimization step, respectively. Despite nonconvexity, we prove that Alt-PGD-Min achieves ϵ-accuracy recovery with O (κ2 log 1ϵ ) iteration complexity and O (κ6rn3 log n3 (κ2r (n1 + n2) + n1 log 1ϵ)) sample complexity, where κ denotes tensor condition number of X★. To further accelerate the convergence, especially when the tensor is ill-conditioned with large κ, we prove AltScalePGD-Min that preconditions the gradient update using an approximate Hessian that can be computed efficiently. We show that Alt-ScalePGD-Min achieves κ independent iteration complexity O(log 1ϵ) and improves the sample complexity to O (κ4rn3 log n3 (κ4r(n1 + n2) + n1 log 1ϵ)). Experiments validate the effectiveness of the proposed methods. © 2025, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Automatic Segmentation of Handwritten Devanagari Word Documents Enabling Accurate Recognition 1

引用

9th International Conference on Mining Intelligence and Knowledge Exploration, MIKE 2021

作者： Bhat, Mohammad Idrees Sharada, B. Imran, Mohammad Obaidullah, Sk. Md School of Computer Science MIT-World Peace University Maharashtra Pune India Department of Studies in Computer Science University of Mysore Karnataka Mysuru India NTT DATA Information Processing Services Private Limited Karnataka Bangalore India Department of Computer Science and Engineering Aliah University West-Bengal India

ISBN: (数字)9783031215179

ISBN: (纸本)9783031215162

In this paper, we propose different approaches for the segmentation of handwritten Devanagari word documents into constituent characters (or pseudo-characters). For accurate identification and segmentation of shiroreakha we exploited ShiroreakhaNet which is encoder-decoder based convolutional neural network. After, segmenting the shiroreakha structural patterns/properties are exploited for the segmentation of upper and lower modifiers. For the corroboration of the efficacy of the results, we collected dataset from different domains. Comparison is also performed with the state-of-the-art methods, and it was revealed that proposed approaches significantly perform better. © 2022, Springer Nature Switzerland AG.

关键词： Character recognition

来源：评论

学校读者我要写书评

暂无评论

Automated and Decentralized Cloud based water level audit system with loT 9

Automated and Decentralized Cloud based water level audit sy...

引用

9th International Conference on science, Technology, Engineering and Mathematics, ICONSTEM 2024

作者： Selvakumar, S. Geetha, C. Muthulakshmi, R. Vidhya Cheerla, Sreevardhan Kumar, Ashok Joshi, Sudhir AarupadiVeedu Institute of Technology Vinayaka Mission Research Foundation Department of Computer Science And Engineering Tamil Nadu Paiyanur603104 India R. M. K. Engineering College Department of Computer Science And Engineering Tamil Nadu Kavaraipettai601206 India Panimalar Engineering College Department of Artificial Intelligence And Data Science Chennai Tamil Nadu Poonamallee600123 India KoneruLakshmaiah Education Foundation Deportmem of Electronics And Communications Engineering Andhra Pradesh Vaddeswaram522302 India Banasthali Vidyapith Department of Computer Science Rajasthan 304022 India Graphic Era Deemed To Be University Department of Mechanical Engineering Uttarakhand Dehradun248002 India

ISBN: (纸本)9798350365092

Water bodies play a fundamental role in the development of society. The construction of dams for water storage has been carried out since ancient civilizations and, along with the development of modern society, new forms of construction have emerged. Over time, these constructions led to improvements for the population which, in turn, took the price of changing the environment and its effects. On this paper a water level audit system was developed with hardware for sensing and sending the data through the cellular network to a server for storage and later displaying it to the user so that he doesn't need to go back to the essay location, being useful for research on the local hydraulic system. © 2024 IEEE.

关键词： Water levels

来源：评论

学校读者我要写书评

暂无评论

Cycle Based Link Prediction Method for Cross-Granularity Contrast Learning

Cycle Based Link Prediction Method for Cross-Granularity Con...

引用

Medical Artificial Intelligence (MedAI), IEEE International Conference on

作者： Zehua Zhang Feihu Zheng Bin Wang Surui Yan Faculty of Computer Science and Technology (School of Big Data) Taiyuan University of Technology Jinzhong China

ISBN: (数字)9798350377613

ISBN: (纸本)9798350377620

In the process of network evolution, interactions be-tween network nodes often extend beyond simple binary interactions to include group interactions. Examples include social networks like those found on Facebook and research collaboration networks formed through scientific partnerships. Link prediction in research collaboration networks aims to forecast collaborative relationships among scientists, thereby enhancing our understanding and opti-mizing the structure and dynamics of these networks. However, net-work data are frequently sparse, and prediction based solely on at-tributes or structural features often yields inconsistent results. In complex networks, higher-order structures often contain more comprehensive information. Capturing the cycle structure in the graph enriches the representation of nodes as well as reduces the infor-mation loss during the information transfer process, thus improving the accuracy of link prediction between nodes. Specifically, our approach utilizes the cycle to calculate its significance and reconstructs ‘fat node’ information, which replaces node representations within the original cycle. We introduce a cycle scale module to de-termine the optimal cycle size for the graph. Moreover, by constructing the clustering coefficients of the multivariate cycle, we identify node and cluster aggregations within the graph, thereby reduce the complexity of finding cycle scales. Subsequently, a graph compari-son learning model is employed to establish a mapping between the newly generated ‘fat nodes’ and link relationships. Through experi-mental validation, our method demonstrates strong link prediction performance across various real network datasets.

关键词： Accuracy Social networking (online) Computational modeling Collaboration Predictive models Prediction algorithms Complexity theory Computational efficiency Convolutional neural networks Optimization

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：