检索结果-内蒙古大学图书馆

2025 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2025

作者： Zhang, Shuailei Ding, Yi Jiang, Muyun Tang, Ning Chew, Effie Ang, Kai Keng Guan, Cuntai College of Computing and Data Science Nanyang Technological University Singapore National University Hospital Singapore Singapore National University Hospital National University of Singapore Singapore Agency for Science Technology and Research Nanyang Technological University Singapore

ISBN: (纸本)9798350368741

Brain-computer interfaces (BCIs) hold great potential for motor recovery in post-stroke patients. However, the motor imagery decoding accuracy is limited by the nonstationarity of EEG signals across subjects and sessions. We propose CAT-Net: a Co-Adaptive Transfer learning network to simultaneously address the inter-subject variability and intersession nonstationarity in EEG data. The proposed method selects a relevant subset of data from all the available subjects' data to train an initial model, followed by subject-specific transfer learning from the initial model to the target subject to establish a pretrain model. Subsequently, online adaptive training is then applied to incrementally train the pretrain model using the data from previous sessions for the target subject. This proposed network using this unique co-adaptive training method is then evaluated on both upper and lower-limb neurorehabilitation EEG datasets comprising 358 sessions from 33 stroke patients. The results showed significant accuracy improvements, achieving averaged accuracies of 70.6% and 72.3% on the respective datasets, surpassing the state-of-the-art baselines. © 2025 IEEE.

关键词： Electroencephalography Neurorehabilitation Transfer Learning

来源：评论

学校读者我要写书评

暂无评论

Localization-Assisted Fast and Robust Beam Optimization for mmWave Communications

引用

IEEE Internet of Things Journal 2025年第9期12卷 12383-12396页

作者： Xiao, Qiqi He, Yinghui Yu, Guanding Yuan, Jiantao Yin, Rui Zhejiang University College of Information Science and Electronic Engineering Hangzhou310027 China Nanyang Technological University College of Computing and Data Science Jurong West 639798 Singapore Hangzhou City University School of Information and Electrical Engineering Academy of Edge Intelligence Hangzhou310015 China Zhengzhou Digital Industry Institute Hangzhou City University Zhengzhou450046 China Hangzhou City University School of Information and Electrical Engineering Hangzhou310015 China

The millimeter wave (mmWave) communication becomes a key enabler for the future Internet of Things (IoT) due to its capability for supporting high rate and low-latency traffic. However, beamforming in the mmWave band faces issues of low efficiency since the narrow beam of mmWave devices would increase the search delay and overhead. Inspired by this, we utilize localization over sub-6 GHz band to assist the mmWave base station in performing fast and robust adaptive beamforming (RABF). Different from existing works, we focus on the indoor scenario and consider the effects of several practical issues, including localization errors and hardware defects. Specifically, a novel two-step access scheme is proposed. During the first step, we design a novel localization method customized for indoor scenarios, jointly considering the time of flight and angle of arrival. The localization error is further analyzed to determine the mmWave scanning angle and an optimal beamwidth expression is derived in closed-form to maximize system throughput with the considerations of the search delay. Moreover, considering the mismatch of the steering vector caused by the hardware defects, we propose an RABF method in closed-form. Simulation results demonstrate that the proposed scheme can effectively reduce the search delay and realize robust beamforming to enhance the mmWave communication performance. © 2014 IEEE.

关键词： Beamforming

来源：评论

学校读者我要写书评

暂无评论

Securing Automated Insulin Delivery Systems: A Review of Security Threats and Protectives Strategies

arXiv

引用

arXiv 2025年

作者： Niu, Yuchen Lam, Siew-Kei College of Computing and Data Science Nanyang Technological University 50 Nanyang Ave 639798 Singapore

Automated insulin delivery (AID) systems have emerged as a significant technological advancement in diabetes care. These systems integrate a continuous glucose monitor, an insulin pump, and control algorithms to automate insulin delivery, reducing the burden of self-management and offering enhanced glucose control. However, the increasing reliance on wireless connectivity and software control has exposed AID systems to critical security risks that could result in life-threatening treatment errors. This review first presents a comprehensive examination of the security landscape, covering technical vulnerabilities, legal frameworks, and commercial product considerations, and an analysis of existing research on attack vectors, defence mechanisms, as well as evaluation methods and resources for AID systems. Despite recent advancements, several open challenges remain in achieving secure AID systems, particularly in standardising security evaluation frameworks and developing comprehensive, lightweight, and adaptive defence strategies. As one of the most widely adopted and extensively studied physiologic closed-loop control systems, this review serves as a valuable reference for understanding security challenges and solutions applicable to analogous medical systems. Copyright © 2025, The Authors. All rights reserved.

关键词： Closed loop control systems

来源：评论

学校读者我要写书评

暂无评论

Efficient Fourier Filtering Network with Contrastive Learning for UAV-based Unaligned Bi-modal Salient Object Detection

引用

IEEE Transactions on Geoscience and Remote Sensing 2025年 63卷

作者： Lyu, Pengfei Yeung, Pak-Hei Yu, Xiaosheng Cheng, Xiufei Wu, Chengdong Rajapakse, Jagath C. Northeastern University Faculty of Robot Science and Engineering Shenyang110169 China Nanyang Technological University College of Computing and Data Science 639798 Singapore

Unmanned aerial vehicle (UAV)-based bi-modal salient object detection (BSOD) aims to segment salient objects in a scene utilizing complementary cues in unaligned RGB and thermal image pairs. However, the high computational expense of existing UAV-based BSOD models limits their applicability to real-world UAV devices. To address this problem, we propose an efficient Fourier filter network with contrastive learning that achieves both real-time and accurate performance. Specifically, we first design a semantic contrastive alignment loss to align the two modalities at the semantic level, which facilitates mutual refinement in a parameter-free way. Second, inspired by the fast Fourier transform that obtains global relevance in linear complexity, we propose synchronized alignment fusion, which aligns and fuses bi-modal features in the channel and spatial dimensions by a hierarchical filtering mechanism. Our proposed model, AlignSal, reduces the number of parameters by 70.0%, decreases the floating point operations by 49.4%, and increases the inference speed by 152.5% compared to the cutting-edge BSOD model (i.e., MROS). Extensive experiments on the UAV RGB-T 2400 and seven bi-modal dense prediction datasets demonstrate that AlignSal achieves both real-time inference speed and better performance and generalizability compared to nineteen state-of-the-art models across most evaluation metrics. In addition, our ablation studies further verify AlignSal's potential in boosting the performance of existing aligned BSOD models on UAV-based unaligned data. The code is available at: https://***/JoshuaLPF/AlignSal. © 2025 IEEE.

关键词： Unmanned aerial vehicles (UAV)

来源：评论

学校读者我要写书评

暂无评论

Multi-Classification Refactoring Framework Using Hopfield Neural Network for Sustainable Software Development

引用

IEEE Access 2025年 13卷 31785-31808页

作者： Almogahed, Abdullah Mahdin, Hairulnizam Hyeon Gu, Yeong Al-Masni, Mohammed A. Alzaeemi, Shehab Abdulhabib Omar, Mazni Alawadhi, Abdulwadood Obaid Barraood, Samera Rehman Gilal, Abdul Ameen Bakather, Adnan Universiti Tun Hussein Onn Malaysia Faculty of Computer Science and Information Technology Parit Raja Johor86400 Malaysia Sejong University College of AI Convergence Department of Artificial Intelligence and Data Science Seoul05006 Korea Republic of King Fahd University of Petroleum and Minerals Interdisciplinary Research Center in Finance and Digital Economy Business School Dhahran31261 Saudi Arabia Universiti Utara Malaysia School of Computing Sintok06010 Malaysia Hadhramout University College of Computers and Information Technology Department of Computer Science Hadhramaut Yemen Florida International University Knight Foundation School of Computing and Information Sciences MiamiFL33199 United States

Adopting high-quality source code is the ultimate way through which software evolution can be ensured as sustainable. Continuous refactoring in complex software systems ensures longevity and increases architecture knowledge sustainability. However, decision-making about refactoring is a challenge because the benefits of refactoring are vague and very difficult for the developers to quantify, as different refactoring strategies have different effects on quality attributes. No research has developed a multi-classification refactoring framework using artificial neural networks (ANN), specifically hopfield neural networks (HNN), to classify refactoring strategies and improve external software quality and sustainability. Therefore, this study proposes a multi-classification refactoring framework using HNN that classifies refactoring strategies by their impact on external quality attributes. Five stages have been conducted to perform this study, including selecting case studies, identifying the external quality attributes, identifying the most commonly used refactoring strategies in practice, conducting the experiments, and conducting the classification process using HNN. The proposed framework categorizes the refactoring strategies into three categories (positive, negative, and ineffective). By providing clear classifications and descriptions of each strategy, the proposed framework helps developers make informed decisions about how to improve the design and structure of their code. It helps developers mitigate risks associated with code changes by providing guidance on which strategies are likely to yield positive results for specific quality attributes. The proposed multi-classification refactoring framework enhances software sustainability by enhancing critical quality attributes. It supports maintainability, adaptability, and long-term viability, helping to ensure that the software systems remain relevant, efficient, and valuable over time. © 2025 IEEE.

关键词： Hopfield neural networks

来源：评论

学校读者我要写书评

暂无评论

Contrastive-Based Removal of Negative Information in Multimodal Emotion Analysis

引用

Cognitive Computation 2025年第3期17卷 1-16页

作者： Wang, Rui Wang, Yaoyang Cambria, Erik Fan, Xuhui Yu, Xiaohan Huang, Yao E, Xiaosong Zhu, Xianxun School of Communication and Information Engineering Shanghai University Shanghai China College of Computing and Data Science Nanyang Technological University Singapore Singapore School of Computing Macquarie University Sydney Australia

Multimodal sentiment analysis bridges the communication gap between humans and machines by accurately recognizing human emotions. However, existing approaches often focus on synchronizing multimodal data to enhance accuracy, overlooking the critical role of negative information. Negative information generally refers to noise or inconsistencies with primary emotional labels within the data, such as disparities in emotional expressions across different modalities and noisy data elements. These issues can significantly compromise the effectiveness of sentiment analysis systems. To address this challenge, we propose a novel method based on contrastive learning for the removal of non-relevant features within single modalities, aiming to eliminate negative information in speech, text, and image data. Additionally, we have designed an enhanced multi-head attention mechanism that integrates the cleansed features into a unified representation for emotion analysis. Experimental evaluations on the CMU-MOSI and CMU-MOSEI datasets demonstrate that our method significantly outperforms existing approaches in sentiment analysis tasks. This method not only improves accuracy but also ensures the system’s robustness against the diverse and noisy nature of real-world data. The relevant code is available at https://***/YaoYangWang/MECAM .

关键词：

来源：评论

学校读者我要写书评

暂无评论

Hop-level Direct Preference Optimization for Knowledge Graph Reasoning with Trees

Hop-level Direct Preference Optimization for Knowledge Graph...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Tiesunlong Shen Jin Wang Xuejie Zhang Erik Cambria School of Information Science and Engineering Yunnan University China College of Computing and Data Science Nanyang Technological University Singapore

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Recent advancements in knowledge graph question answering (KGQA) have shown promise, yet existing methods often fail to align with human reasoning patterns. This study proposes HD-PORT (hop-level direct preference optimization for knowledge graph reasoning with trees), a novel approach that combines Monte Carlo Tree Search (MCTS) with hop-level direct preference optimization (HDPO) for KGQA tasks. MCTS simulates human-like reasoning by exploring multiple inference paths in knowledge graphs, generating interpretable reasoning chains and rich hop-level preferences. HDPO then leverages these preferences for model optimization, addressing the limitations of traditional supervised learning and existing DPO methods that focus on overall solution preferences. By optimizing preferences at each reasoning step, HD-PORT more closely mirrors human problem-solving strategies. Experimental results on benchmark datasets demonstrate that HD-PORT significantly outperforms state-of-the-art methods in both accuracy and interpretability, particularly for complex, multi-hop reasoning tasks.

关键词： Monte Carlo methods Supervised learning Knowledge graphs Signal processing Cognition Question answering (information retrieval) Problem-solving Mirrors Speech processing Optimization

来源：评论

学校读者我要写书评

暂无评论

Blockchain Integration for Robust Intellectual Property Protection in New Product Development (NPD) 3

Blockchain Integration for Robust Intellectual Property Prot...

引用

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

作者： Chinnasamy, P. Subashini, B. Nijanthan, Nijanthan Madasamy Raja, G. Swetha, Devulapally Sreenivasulu Reddy, L. School of Computing Kalasalingam Academy of Research and Education Department of Computer Science and Engineering Tamilnadu India School of Computing Srmist Department of Data Science and Business Systems Kattankulathur India Paavai Engineering College Tamilnadu Namakkal India Paavai Engineering College Tamilnadu Namakkal India Vaagdevi Engineering College Dept. of Computer Science and Engineering Warangal Dist. Telangana Bolli Kunta506005 India School of Advanced Sciences Kalasalingam Academy of Research and Education Department of Mathematics Tamilnadu Krishnankoil India

ISBN: (纸本)9798331509675

The rapid advancement of Blockchain technology has led to its widespread application in several domains of digital activities, such as e-government concerns and monetary security. This article presents a smart contract-based Blockchain solution for protecting intellectual property rights in new product development. The proposed solution effectively tackles the challenges associated with the confidentiality of personally identifiable information, efficiency in storing transaction-related information, as well as cooperation and monitoring in IPR. Initially, we create a mathematical model based on smart contract enabled Blockchain and intelligent contracts. Subsequently, we develop algorithms for Blockchain generation, and intelligent contracts. The system is suitable for protecting IPR in a blockchain network, by implementation of distributed identity authentication and real-time extensible storage. These features enhance the security and expandability of circuit copyright protection based on blockchain technology. The experimental results demonstrate that the suggested algorithm has decreased the cost for communication and enhanced the effectiveness of storing information and administration. Furthermore, it demonstrates resilience against many common attacks such as double-spending assaults, while maintaining low cost and overhead. Additionally, it exhibits a superior level of security in comparison to three other competing algorithms. © 2025 IEEE.

关键词： Smart contract

来源：评论

学校读者我要写书评

暂无评论

Fuzzy-Based Optimization Techniques for Segmenting the Tumors in Multimodal MRI Images

引用

Operations Research Forum 2025年第1期6卷 1-27页

作者： Alagarsamy, Saravanan Nagarajan, D. Govindaraj, Vishnuvardhan Department of Information Technology Sri Sivasubramaniya Nadar College of Engineering Rajiv Gandhi Salai (OMR) Tamilnadu Chennai India Department of Mathematics Rajalakshmi Institute of Technology Chennai India Division of Data Science of School of Computing Science and Education VIT Bhopal University Kothrikalan Madhya Pradesh Sehore 466114 India

Accurate brain tumor prediction is essential when approaching the field of health care;wherever accuracy in decision-making is crucial, the issues also need to be resolved right away. Many artificial intelligence and machine learning-based techniques have been developed in the last couple of years in the field of healthcare. The intention of this work is to create an algorithm that combines the features of Crow Search (CS) and Interval Type-II Fuzzy Logic System (IT2FLS) algorithms to distinguish the area of tumor from complex brain tissues. The ability of oncologists to make decisions is critical for any therapy sequence to be successful, and the methodology proposed in the research work considerably influences a conclusion through technology interruption. The suggested approach is flexible enough to work with a diversity of image sequences found in the BRATS Challenge 2020 dataset that presents varying degrees of obstacles, challenges, and difficulties in locating the anomalous regions, and it produces improved demarcation results that have been instinctively evaluated and supported by Dice score (98 ± 1.3), specificity (98 ± 1.4), and sensitivity (98 ± 1.7) as average metrics. The motive of this study is the enhancement of the visual perception of oncologists, which gives them improved insight into and comprehension of the condition of the patient. © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2025.

关键词： Crow Search (CS) Interval Type-II Fuzzy Logic System (IT2FLS) Magnetic Resonance Imaging (MRI) Tumor segmentation

来源：评论

学校读者我要写书评

暂无评论

Towards Label-Only Membership Inference Attack against Pre-trained Large Language Models

arXiv

引用

arXiv 2025年

作者： He, Yu Li, Boheng Liu, Liu Ba, Zhongjie Dong, Wei Li, Yiming Qin, Zhan Ren, Kui Chen, Chun The State Key Laboratory of Blockchain and Data Security Zhejiang University China College of Computing and Data Science Nanyang Technological University Singapore

Membership Inference Attacks (MIAs) aim to predict whether a data sample belongs to the model’s training set or not. Although prior research has extensively explored MIAs in Large Language Models (LLMs), they typically require accessing to complete output logits (i.e., logits-based attacks), which are usually not available in practice. In this paper, we study the vulnerability of pre-trained LLMs to MIAs in the label-only setting, where the adversary can only access generated tokens (text). We first reveal that existing label-only MIAs have minor effects in attacking pre-trained LLMs, although they are highly effective in inferring fine-tuning datasets used for personalized LLMs. We find that their failure stems from two main reasons, including better generalization and overly coarse perturbation. Specifically, due to the extensive pretraining corpora and exposing each sample only a few times, LLMs exhibit minimal robustness differences between members and non-members. This makes token-level perturbations too coarse to capture such differences. To alleviate these problems, we propose PETAL: a label-only membership inference attack based on PEr-Token semAntic simiLarity. Specifically, PETAL leverages token-level semantic similarity to approximate output probabilities and subsequently calculate the perplexity. It finally exposes membership based on the common assumption that members are ‘better’ memorized and have smaller perplexity. We conduct extensive experiments on the WikiMIA benchmark and the more challenging MIMIR benchmark. Empirically, our PETAL performs better than the extensions of existing label-only attacks against personalized LLMs and even on par with other advanced logit-based attacks across all metrics on five prevalent open-source LLMs1. Our study highlights that pre-trained LLMs remain vulnerable to privacy risks posed by MIAs even in the most challenging and realistic setting, calling for attention to develop more effective defenses. Copyright © 20

关键词： Semantics

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：