检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Lawrence, Hannah Portilheiro, Vasco Zhang, Yan Kaba, Sékou-Oumar Department of Electrical Engineering and Computer Science MIT United States Gatsby Computational Neuroscience Unit UCL United Kingdom Samsung – SAIT AI Lab Montreal Canada Mila – Quebec Artficial Intelligence Institute Canada McGill University Canada

Equivariance encodes known symmetries into neural networks, often enhancing generalization. However, equivariant networks cannot break symmetries: the output of an equivariant network must, by definition, have at least the same self-symmetries as the input. This poses an important problem, both (1) for prediction tasks on domains where self-symmetries are common, and (2) for generative models, which must break symmetries in order to reconstruct from highly symmetric latent spaces. This fundamental limitation can be addressed by considering equivariant conditional distributions, instead of equivariant functions. We present novel theoretical results that establish necessary and sufficient conditions for representing such distributions. Concretely, this representation provides a practical framework for breaking symmetries in any equivariant network via randomized canonicalization. Our method, SymPE (Symmetry-breaking Positional Encodings), admits a simple interpretation in terms of positional encodings. This approach expands the representational power of equivariant networks while retaining the inductive bias of symmetry, which we justify through generalization bounds. Experimental results demonstrate that SymPE significantly improves performance of group-equivariant and graph neural networks across diffusion models for graphs, graph autoencoders, and lattice spin system modeling. © 2025, CC BY-NC-SA.

关键词： Graph neural networks

来源：评论

学校读者我要写书评

暂无评论

Flow-TSVAD: Target-Speaker Voice Activity Detection via Latent Flow Matching for Speaker Diarization

Flow-TSVAD: Target-Speaker Voice Activity Detection via Late...

引用

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

作者： Zhengyang Chen Bing Han Shuai Wang Yidi Jiang Yanmin Qian Department of Computer Science and Engineering Auditory Cognition and Computational Acoustics Lab MoE Key Lab of Artificial Intelligence AI Institute Shanghai Jiao Tong University Shanghai China Shenzhen Research Institute of Big Data Shenzhen China School of Data Science The Chinese University of Hong Kong Shenzhen China National University of Singapore Singapore

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Speaker diarization is typically considered as a discriminative task, using discriminative approaches to produce fixed diarization results. In this paper, we explore for the first time the use of neural network-based generative methods for speaker diarization. We implement a Flow-Matching (FM) based generative algorithm within the sequenceto-sequence target speaker voice activity detection (Seq2Seq-TSVAD) diarization system. Our experiments reveal that applying the generative method directly to the original binary label sequence space of the TS-VAD output is ineffective. To address this issue, we propose mapping the binary label sequence into a dense latent space before applying the generative algorithm, and our proposed Flow-TSVAD method can significantly outperform the traditional Seq2Seq-TSVAD system. Additionally, we observe that the FM algorithm converges rapidly during the inference stage, only requiring two inference steps to achieve promising results. Moreover, as a generative model, Flow-TSVAD allows for sampling different diarization results by running the model multiple times, so the ensemble system combining the results from various sampling instances can further boost the diarization performance.

关键词： Voice activity detection Frequency modulation computational modeling Signal processing algorithms Signal processing Inference algorithms Acoustics computational efficiency

来源：评论

学校读者我要写书评

暂无评论

Evaluating Cyber-Crime Using Machine Learning and AI Approach for Environmental Sustainability 2nd

Evaluating Cyber-Crime Using Machine Learning and AI Approac...

引用

2nd International Conference on Sustainable Development Using Machine Learning, Artificial Intelligence, and IoT, ICSD 2024

作者： Chauhan, Ritu Mehtar, Khushi Kaur, Harleen Alankar, Bhavya Artificial Intelligence and IOT Lab Center for Computational Biology and Bioinformatics Amity University Uttar Pradesh Noida India Amity Institute of Biotechnology Amity University Uttar Pradesh Noida India Department of Computer Science and Engineering School of Engineering Sciences and Technology Jamia Hamdard New Delhi India

ISBN: (纸本)9783031717284

The online world has become an essential part of everyday life in India due to its fast expansion and technical improvements. But as technology has advanced, cyber-crimes have also increased dramatically, posing serious risks to the economy, security, and privacy. This research examines the intricate terrain of cybercrimes targeting minors in India, scrutinizing more than 16 lakh cases that have been documented during the three years preceding this one. The study takes an extensive approach, applying machine learning modelling such as Gradient Boosting, AdaBoost, Random Forest, Gaussian Naive Bayes, Bernoulli Naive Bayes, Linear SVM, Logistic Regression, and Gradient Boosting in conjunction with exploratory assessment of data. The findings show different accuracy rates;Random Forest, Gradient Boosting, Linear SVM, and Gaussian Naive Bayes all show an impressive 88% accuracy. AdaBoost and Logistic Regression reach 75%, but Bernoulli Naive Bayes comes in lower at 62%. The comprehension of computational interconnections is improved with correlation matrices. This thorough analysis provides important insights for academics, policymakers, and stakeholders in the field of child cybersecurity by exposing common cyber threats targeting minors and highlighting the role that algorithms for learning may play in detecting and averting these risks for environmental sustainability. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

DEUCE: Dual-diversity Enhancement and Uncertainty-awareness for Cold-start Active Learning

arXiv

引用

arXiv 2025年

作者： Guo, Jiaxin Chen, C.L. Philip Li, Shuzhen Zhang, Tong Guangdong Provincial Key Laboratory of Computational AI Models and Cognitive Intelligence School of Computer Science and Engineering South China University of Technology Guangzhou China Pazhou Lab Guangzhou China Engineering Research Center of the Ministry of Education on Health Intelligent Perception and Paralleled Digital-Human Guangzhou China

Cold-start active learning (CSAL) selects valuable instances from an unlabeled dataset for manual annotation. It provides high-quality data at a low annotation cost for label-scarce text classification. However, existing CSAL methods overlook weak classes and hard representative examples, resulting in biased learning. To address these issues, this paper proposes a novel dual-diversity enhancing and uncertainty-aware (DEUCE) framework for CSAL. Specifically, DEUCE leverages a pre-trained language model (PLM) to efficiently extract textual representations, class predictions, and predictive uncertainty. Then, it constructs a Dual-Neighbor Graph (DNG) to combine information on both textual diversity and class diversity, ensuring a balanced data distribution. It further propagates uncertainty information via density-based clustering to select hard representative instances. DEUCE performs well in selecting class-balanced and hard representative data by dual-diversity and informativeness. Experiments on six NLP datasets demonstrate the superiority and efficiency of DEUCE. © 2025, CC BY.

关键词： Active learning

来源：评论

学校读者我要写书评

暂无评论

FACE: Few-shot Adapter with Cross-view Fusion for Cross-subject EEG Emotion Recognition

arXiv

引用

arXiv 2025年

作者： Liu, Haiqi Chen, C.L. Philip Zhang, Tong Guangdong Provincial Key Laboratory of Computational AI Models and Cognitive Intelligence School of Computer Science and Engineering South China University of Technology Guangzhou510006 China Pazhou Lab Guangzhou510335 China Engineering Research Center Ministry of Education on Health Intelligent Perception and Paralleled Digital-Human Guangzhou China

Cross-subject EEG emotion recognition is challenged by significant inter-subject variability and intricately entangled intra-subject variability. Existing works have primarily addressed these challenges through domain adaptation or generalization strategies. However, they typically require extensive target subject data or demonstrate limited generalization performance to unseen subjects. Recent few-shot learning paradigms attempt to address these limitations but often encounter catastrophic overfitting during subject-specific adaptation with limited samples. This article introduces the few-shot adapter with a cross-view fusion method called FACE for cross-subject EEG emotion recognition, which leverages dynamic multi-view fusion and effective subject-specific adaptation. Specifically, FACE incorporates a cross-view fusion module that dynamically integrates global brain connectivity with localized patterns via subject-specific fusion weights to provide complementary emotional information. Moreover, the few-shot adapter module is proposed to enable rapid adaptation for unseen subjects while reducing overfitting by enhancing adapter structures with meta-learning. Experimental results on three public EEG emotion recognition benchmarks demonstrate FACE’s superior generalization performance over state-of-the-art methods. FACE provides a practical solution for cross-subject scenarios with limited labeled data. Copyright © 2025, The Authors. All rights reserved.

关键词： Emotion Recognition

来源：评论

学校读者我要写书评

暂无评论

Full-Waveform Inversion With Low-frequency Extrapolation Based on Sparse Deconvolution for Ultrasound Computed Tomography

引用

Ultrasound in Medicine and Biology 2025年

作者： Zhang, Nuomin Xiao, Yang Yuan, Yu Yang, Xudong Shen, Yi Department of Control Science and Engineering Harbin Institute of Technology Harbin China Computational Sensing Research Center Zhejiang lab Hangzhou China Zhejiang Equilibrium Nine Medical Technology Co. Ltd Zhejiang Hangzhou China

Objective: In ultrasound computed tomography (USCT), full-waveform inversion (FWI) is a promising algorithm for high-resolution sound-speed reconstruction. When implementing FWI in practical imaging systems, insufficient high-quality, low-frequency information often leads to cycle skipping, which subsequently degrades convergence and accuracy. To address this problem, this paper proposes a modified FWI algorithm. Methods: Our approach incorporated low-frequency extrapolation for seismic imaging applications, capitalizing on the inherent sparsity of time-domain impulse response functions. Through a deconvolution-based framework, we enabled robust impulse response function estimation that facilitated the spectral extension of band-limited measurements. The extrapolated low-frequency components, while representing an approximate recovery rather than exact reconstruction of unmeasured frequencies, demonstrated sufficient fidelity for practical implementation in multi-frequency inversion workflows. Results: Numerical and experimental studies have demonstrated the efficacy of extrapolated low-frequency components in mitigating cycle-skipping artifacts. Compared with conventional low-pass filtering, the proposed method reduced the sound-speed reconstruction root mean square error from 34.47 m/s to 6.47 m/s. Phantom experiments confirmed the robustness of our method, demonstrating root mean square error reduction from 16.57 m/s (standard filtering) to 5.98 m/s (our method). Conclusion: This work relaxes the restriction of FWI in transducer frequency, potentially making FWI more compatible with high-frequency imaging modalities. © 2025 World Federation for Ultrasound in Medicine & Biology

关键词： Image coding

来源：评论

学校读者我要写书评

暂无评论

Markerless Sagittal Gait Analysis of Cerebral Palsy in Children Using Pose Estimation Techniques

Markerless Sagittal Gait Analysis of Cerebral Palsy in Child...

引用

Innovation in Computing and engineering (ICE), International Conference on

作者： Vinay Kumar Vats Chandra Prakash Amandeep Kaur Rajesh Kumar Computer Science and Engineering National Institute of Technology Delhi India Computational Intelligence and Smart Motion Research (CISMR) Lab SVNIT Surat India Malaviya National Institute of Technology Jaipur India Department of Human Anatomy and Physiology Faculty of Health Sciences University of Johannesburg Johannesburg South Africa

ISBN: (数字)9798350380460

ISBN: (纸本)9798350380477

Cerebral Palsy (CP) is a neurological disorder affecting children’s motor function and posture. Early detection and monitoring of Cerebral palsy is crucial for effective intervention and improved quality of life. Gait analysis is a critical tool for assessing motor function in children with cerebral palsy, often requiring controlled environments and markers for accuracy. Markerless gait analysis has emerged as a promising tool for objectively evaluating gait abnormalities in children with CP. In this study, we propose a markerless methodology utilising pose estimation to analyse gait patterns from sagittal video recordings captured in unconstrained environments. The dataset consists of 49 children, including 10 typically developing children and 39 with cerebral palsy, focusing on side-view videos for the analysis. Given the motor disabilities of these children, conducting gait analysis in a constrained environment can alter their natural walking patterns, making our approach particularly advantageous. Then we identify the key gait events, heel strike and toe-off, to distinguish gait between cycles by tracing knee joint movements. This technique demonstrates the potential for practical, non-invasive gait assessment in natural settings, offering insights into the distinct gait characteristics of children with cerebral palsy compared to their typically developing peers.

关键词： Legged locomotion Neurological diseases Cerebral palsy Technological innovation Pose estimation Motors Motion control Gait recognition Video recording Monitoring

来源：评论

学校读者我要写书评

暂无评论

The role of prior in image based 3D modeling： a survey

引用

Frontiers of Computer science 2017年第2期11卷 175-191页

作者： Hao ZHU Yongming NIE Tao YUE Xun CAO Lab for Computational Imaging Technology and Engineering School of Electronic Science and Engineering Nanjing University Nanjing 210023 China

The prior knowledge is the significant supple- ment to image-based 3D modeling algorithms for refining the fragile consistency-based stereo. In this paper, we review the image-based 3D modeling problem according to prior cate- gories, i.e., classical priors and specific priors. The classical priors including smoothness, silhouette and illumination are well studied for improving the accuracy and robustness of the 3D reconstruction. In recent years, various specific priors which take advantage of Manhattan rule, geometry template and trained category features have been proposed to enhance the modeling performance. The advantages and limitations of both kinds of priors are discussed and evaluated in the paper. Finally, we discuss the trend and challenges of the prior studies in the future.

关键词： prior information consistency-based stereo,smoothness illumination silhouette specific prior

来源：评论

学校读者我要写书评

暂无评论

MHD analysis of couple stress nanofluid through a tapered non-uniform channel with porous media and slip-convective boundary effects

International Journal of Thermofluids

引用

International Journal of Thermofluids 2025年 27卷

作者： Deepalakshmi, P. Shankar, G. Siva, E.P. Tripathi, D. Bég, O. Anwar Department of Mathematics College of Engineering and Technology SRM Institute of Science and Technology Tamil Nadu Kattankulathur603203 India Centre for Computational Biology Easwari Engineering College Tamil Nadu Chennai600089 India Center for Research SRM TRP Engineering College Tamil Nadu Tiruchirappalli621105 India Department of Mathematics National Institute of Technology Uttarakhand 246174 India Multi-Physical Engineering Sciences Group Dept. Mechanical and Aeronautical Engineering Salford University Corrosion/Coatings Lab 3-08 SEE Building Manchester M54WT United Kingdom

The current research addresses the peristaltic transport mechanism that propels fluid through a conduit through rhythmic contraction and relaxation of the conduit walls, a phenomenon evident in numerous biological systems, including the gastrointestinal tract. Motivated by applications in nano-pharmacological drug delivery and thermo-biomagnetic therapy, a mathematical and computational analysis of radiative heat transfer in peristaltic pumping of a magnetohydrodynamic (MHD) couple stress nanofluid through a tapered asymmetric passage, with the influences of a porous medium and wall slip, is presented. Buongiorno's two-component nanoscale model is deployed and the Stokes couple stress non-Newtonian model utilized. Physically the porous medium is modelled with a drag force formulation and simulates the presence of obstructions and deposits in the gastric tract and blood vessels. The governing equations for the couple stress nanofluid are reduced by employing the long-wavelength approximation and the low Reynolds number condition, both standard approaches in fluid dynamics research. Analytical solutions are derived for axial velocity, temperature profile, nanoparticle concentration, stream function, and pressure gradient, providing a comprehensive understanding of the flow dynamics. Furthermore, numerical integration methods are utilized to calculate the average pressure increase (ΔP) and the heat transfer coefficient (Z). The impact of critical parameters namely the Hartmann number (M), Brownian motion parameter (Nb), thermophoresis parameter (Nt), Prandtl number (Pr), slip parameter (L) and radiation parameter (Rn) on fluid dynamics is examined through comprehensive graphical representations. The findings indicate that peristaltic pumping efficiency is superior in a uniform channel relative to a non-uniform channel, underscoring the influence of channel geometry on flow performance. Moreover, the synergistic effects of thermophoresis and Brownian motion result in a

关键词： Reynolds number

来源：评论

学校读者我要写书评

暂无评论

Efficacy of using a digital health intervention model using community health workers for primary health services in Bangladesh: a repeated cross-sectional observational study

引用

BMC public health 2025年第1期25卷 1833页

作者： Marzia Zaman Rubaiyat Alim Hridhee Refat Ahmed Bhuiyan Charles Aunkan Gomes Md Mashiar Rahman Sheikh Mohammed Shariful Islam Farhana Sarker Khondaker A Mamun CMED Health Limited Dhaka Bangladesh. AIMS Lab Institute of Research Innovation Incubation and Commercialization (IRIIC) United International University Dhaka Bangladesh. Department of Computer Science and Engineering United International University Dhaka Bangladesh. Palli Karma-Sahayak Foundation Dhaka Bangladesh. Institute for Physical Activity and Nutrition Faculty of Health Deakin University Melbourne Australia. Center for Computational & Data Sciences (CCDS) and Department of Computer Science and Engineering Independent University Dhaka Bangladesh. AIMS Lab Institute of Research Innovation Incubation and Commercialization (IRIIC) United International University Dhaka Bangladesh. mamun@cse.uiu.ac.bd. Department of Computer Science and Engineering United International University Dhaka Bangladesh. mamun@cse.uiu.ac.bd.

OBJECTIVE:This study aims to evaluate the effectiveness of a digital health intervention model by observing people's health conditions in a rural area of Bangladesh. Through a repeated cross-sectional design, health outcomes were assessed at six-month intervals over 18 months. METHODS:The data presented in this paper are obtained from the model's pilot implementation in one catchment area in Bangladesh. The model involved community health workers (CHWs) using a smart health kit and Artificial Intelligence(AI)-based mobile application to provide monthly doorstep health education, screenings, risk assessments, and digital referrals. Socio-demographic and health measurement data are presented as proportions with 95% confidence intervals (CIs). Multivariate logistic regression was used to analyze the association between diseases and their respective risk factors. We compared health vitals across three consecutive periods to determine the effectiveness of the model in improving health outcomes over time. RESULTS:The model served 32,581 people from 7,090 households during this operation. We found that 21.76% of the served population were overweight, 8.18% had prehypertension, 16.45% had high blood glucose, and 11% children were malnourished. From the analysis of risk factors, we found that people aged were associated with developing hypertension, diabetes, and cardiovascular diseases(CVD). CVD was associated with hypertension and stroke. A comparative analysis of different periods showed an improvement in BMI, BP, and MUAC. Blood glucose measurements did not show significant improvement. CONCLUSION:This study highlights the feasibility and effectiveness of digital health interventions in enhancing primary healthcare in rural settings. Integration of AI-driven decision support with CHW-led health education and screenings can improve early detection, intervention and management of non-communicable diseases (NCDs). Scaling up this model through public-private partnerships o

关键词： Artificial intelligence Clinical decision support system (CDSS) Community health workers (CHWs) Digital health inclusion Digital health intervention model Health equity Non-communicable diseases (NCD) Primary healthcare Sustainable development goals Universal health coverage

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：