检索结果-内蒙古大学图书馆

5th International Conference on Telecommunications, Optics, and computer Science, TOCS 2024

作者： Jia, Chengjie Fei, Rong Li, Xiaomeng School of Computer Science and Engineering Xi'an University of Technology Shaanxi Xi'an 710004 China Shaanxi Key Laboratory for Network Computing and Security Technology School of Computer Science and Engineering Xi'an University of Technology Shaanxi Xi'an 710048 China

ISBN: (纸本)9781510691667

In recent years, with the continuous progress of deep learning neural networks and the construction of large data sets, new breakthroughs in voiceprint verification technology for voice calls have been rapidly achieved. Deep learning-based neural networks, such as X-Vectors, ResNets, Ecapa-Tdnn, etc., are able to utilize their own network characteristics to extract rich voiceprint features, thus further improving the accuracy of voiceprint recognition, but there is still room for improvement in the performance and parameter count of the traditional models. In this paper, we propose the Ecapa2Tdnn model based on EcapaTdnn and ResNet, which has the advantage of possessing a deeper network architecture while performing better in terms of temporal modeling capability. This indicates that the model can effectively handle long time series of speech data, thus improving the performance of the voiceprint recognition system. The experimental results show that under the same experimental environment, compared with the equal error rate (EER) of 11.31% of the EcapaTdnn model, the Ecapa2Tdnn model proposed in this paper has only an EER of 10.27%, which is 1.04% lower than the former EER, and the number of parameters is still one-fourth of the original model. The model improves accuracy while reducing the total number of parameters, which is expected to be widely used in the field of voiceprint recognition. © 2025 SPIE.

关键词： Deep neural network EcapaTdnn ResNet voiceprint recognition model

来源：评论

学校读者我要写书评

暂无评论

Across-Platform Detection of Malicious Cryptocurrency Accounts via Interaction Feature Learning

引用

IEEE Transactions on Information Forensics and Security 2025年 20卷 4783-4798页

作者： Che, Zheng Shen, Meng Tan, Zhehui Du, Hanbiao Wang, Wei Chen, Ting Zhao, Qinglin Xie, Yong Zhu, Liehuang Beijing Institute of Technology School of Computer Science Beijing100081 China Beijing Institute of Technology School of Cyberspace Science and Technology Beijing100081 China Xi'an Jiaotong University Key Laboratory for Intelligent Networks and Network Security Ministry of Education Xi'an710049 China University of Electronic Science and Technology of China School of Computer Science and Engineering Chengdu611731 China Macau University of Science and Technology School of Computer Science and Engineering Taipa China Qinghai Institute of Technology School of Computer and Information Science Xining810016 China

With the rapid evolution of Web3.0, cryptocurrency has become a cornerstone of decentralized finance. While these digital assets enable efficient and borderless financial transactions, their pseudonymous nature has also attracted malicious activities such as money laundering, fraud, and other financial crimes. Effective detection of malicious accounts is crucial to maintaining the security and integrity of the Web 3.0 ecosystem. Existing malicious account detection methods rely on large amounts of labeled data and suffer from low generalization. Label-efficient and generalizable malicious account detection remains a challenging task. In this paper, we propose ShadowEyes, a framework for detecting malicious accounts by leveraging interaction feature learning with only a small labeled dataset. Specifically, We first propose a generalized account representation named TxGraph, which captures the universal interaction features of Ethereum and Bitcoin. Then we carefully design an account representation augmentation method tailored to simulate the evolution of malicious accounts to generate positive pairs. We conduct extensive experiments using public datasets to evaluate the performance of ShadowEyes. The results demonstrate that it outperforms state-of-the-art (SOTA) methods in four typical scenarios. Specifically, in the scenario of across-platform malicious account detection, ShadowEyes maintains an F1 score of around 90%, which is 10% higher than the SOTA method. In the zero-shot learning scenario, it can achieve an F1 score of 79.56% for detecting gambling accounts, surpassing the SOTA method by 10.44%. © 2005-2012 IEEE.

关键词： Zero-shot learning

来源：评论

学校读者我要写书评

暂无评论

A Malicious Websites Classifier Based on an Improved Relation network 6th

A Malicious Websites Classifier Based on an Improved Relatio...

引用

6th International Conference on Machine Learning for Cyber Security, ML4CS 2024

作者： Wang, Qianshi Xu, Chongjun Yang, Huayu Zhai, Xilin Zhang, Hua State Key Laboratory of Networking and Switching Technology Beijing University of Posts and Telecommunications Beijing100876 China Zhejiang Branch of National Computer Network Emergency Response Technical Team/Coordination Center of China Hangzhou China

ISBN: (纸本)9789819645657

Intelligent in-car operating system integrates various web platforms including browsers, therefore facing serious issues of malicious web content. Current detection models perform poorly in Internet of Vehicles (IoV) environment due to simple model structures and lack of data samples, which make models difficult to learn and extract multi-level and fine-grained sample features. In this paper, we propose a novel few-shot model, CarMaNet, to detect and classify malicious websites in IoV environment based on website images. We import inception module into Relation networks, which enhance the ability of CarMaNet to learn multi-level features of images. At the same time, we design a feature pipeline featuring adaptive recalibration between the embedding module and the relation module based on adaptive attention mechanism (FCAA), enhancing the model's ability to learn fine-grained features. In few-shot experiments on three dataset of pornography, gambling and counterfeit websites, CarMaNet demonstrates greater performance than current models. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Websites

来源：评论

学校读者我要写书评

暂无评论

NeuVSA: A Unified and Efficient Accelerator for Neural Vector Search 31

NeuVSA: A Unified and Efficient Accelerator for Neural Vecto...

引用

31st IEEE International Symposium on High Performance computer Architecture, HPCA 2025

作者： Yuan, Ziming Dai, Lei Li, Wen Zhang, Jie Liang, Shengwen Wang, Ying Liu, Cheng Li, Huawei Li, Xiaowei Guo, Jiafeng Wang, Peng Chen, Renhai Zhang, Gong Shanxi University School of Computer and Information Technology China Peking University School of Computer Science China Institute of Computing Technology CAS State Key Lab of Processors China University of Chinese Academy of Sciences China Zhongguancun National Laboratory Beijing China Institute of Computing Technology CAS Key Lab of Network Data Science and Technology China Huawei Technologies Co. Ltd. China

ISBN: (纸本)9798331506476

Neural Vector Search (NVS) has exhibited superior search quality over traditional key-based strategies for information retrieval tasks. An effective NVS architecture requires high recall, low latency, and high throughput to enhance user experience and cost-efficiency. However, implementing NVS on existing neural network accelerators and vector search accelerators is sub-optimal due to the separation between the embedding stage and vector search stage at both algorithm and architecture levels. Fortunately, we unveil that Product Quantization (PQ) opens up an opportunity to break separation. However, existing PQ algorithms and accelerators still focus on either the embedding stage or the vector search stage, rather than both simultaneously. Simply combining existing solutions still follows the beaten track of separation and suffers from insufficient parallelization, frequent data access conflicts, and the absence of scheduling, thus failing to reach optimal recall, latency, and throughput. To this end, we propose a unified and efficient NVS accelerator dubbed NeuVSA based on algorithm and architecture co-design philosophy. Specifically, on the algorithm level, we propose a learned PQ-based unified NVS algorithm that consolidates two separate stages into the same computing and memory access paradigm. It integrates an end-to-end joint training strategy to learn the optimal codebook and index for enhanced recall and reduced PQ complexity, thus achieving smoother acceleration. On the architecture level, we customize a homogeneous NVS accelerator based on the unified NVS algorithm. Each sub-accelerator is optimized to exploit all parallelism exposed by unified NVS, incorporating a structured index assignment strategy and an elastic on-chip buffer to alleviate buffer conflicts for reduced latency. All sub-accelerators are coordinated using a hardware-aware scheduling strategy for boosted throughput. Experimental results show that the joint training strategy improves recall

关键词： network embeddings

来源：评论

学校读者我要写书评

暂无评论

A Byzantine Client Detection Method for Federated Learning Based on Encrypted Repeated Median Linear Regression 24

A Byzantine Client Detection Method for Federated Learning B...

引用

5th International Conference on computer Science and Management technology, ICCSMT 2024

作者： Guo, Jinlong Yang, Ming Li, Huijuan Wu, Xiaoming He, Yunpeng Wang, Xin Key Laboratory of Computing Power Network and Information Security Ministry of Education Shandong Computer Science Center Qilu University of Technology Shandong Academy of Sciences Shandong Jinan China Shandong Evayinfo Technology Co. Ltd. Shandong Jinan China

ISBN: (纸本)9798400709999

Federated learning is an emerging distributed machine learning technology that allows participants to jointly train machine learning models locally without the need for large-scale data transmission and sharing, thereby greatly improving data security. However, some Byzantine clients may intentionally tamper with local model parameter updates with the intention of manipulating the global model's update direction, which can lead to a decline in the overall performance of federated learning. Besides, existing Byzantine client detection research is conducted in plain text, which will lead to the leakage of private information. To solve the above issues, we propose a Byzantine client detection method for federated learning based on encrypted repeated median linear regression. Specifically, it combines the repeated median linear regression algorithm with homomorphic encryption, which can detect and eliminate clients with Byzantine behavior in the ciphertext state, ensuring the security and reliability of the parameters in the entire federated learning process. Additionally, we also designed an adaptive gradient clipping method based on historical iteration to clip the gradient to a reasonable range, which can not only reduce communication and computing overhead, but also improve the convergence speed and stability of the model. Experiments show that the proposed method can maintain high accuracy under the influence of malicious behavior of Byzantine clients. © 2024 Copyright held by the owner/author(s). Publication rights licensed to ACM.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Brillouin Frequency Shift Extraction for BOTDA Based on Gramian Angular Fields 4

Brillouin Frequency Shift Extraction for BOTDA Based on Gram...

引用

4th International Computational Imaging Conference, CITA 2024

作者： Zhu, Borong Zhang, Zhihao Qian, Yuhao Ma, Xiaole Wang, Liang Yu, Kuanglu Beijing Key Laboratory of Advanced Information Science and Network Technology Beijing100044 China Institute of Information Science School of Computer and Information Technology Beijing Jiaotong University Beijing100044 China National Engineering Laboratory for Next Generation Internet Access System School of Optics and Electronic Information Huazhong University of Science and Technology Wuhan430074 China

ISBN: (纸本)9781510688834

To accurately extract Brillouin frequency shift of BOTDA with large sweeping step sizes, a novel structure of GAFCNN is proposed, combining time series coding with convolutional neural networks. The experimental data validation results demonstrate that the GAFCNN is capable of obtaining more accurate results under large sweep steps and low signal-to-noise ratios (SNRs) than other current methods. © 2025 SPIE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

HiRes: Hierarchical Feature Optimization and Rescorer for Automatic ICD Coding

HiRes: Hierarchical Feature Optimization and Rescorer for Au...

引用

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

作者： Liang, Zhenpeng Guan, Hongjiao Lu, Wenpeng Peng, Xueping Xu, Bing Yang, Muyun Key Laboratory of Computing Power Network and Information Security Ministry of Education Shandong Computer Science Center National Supercomputer Center in Jinan Qilu University of Technology Shandong Academy of Sciences Jinan China Shandong Provincial Key Laboratory of Computing Power Internet and Service Computing Shandong Fundamental Research Center for Computer Science Jinan China Australian Artificial Intelligence Institute University of Technology Sydney Sydney Australia Faculty of Computing Harbin Institute of Technology Harbin China

ISBN: (纸本)9798350368741

The International Classification of Diseases (ICD) coding assigns standardized codes to diseases. Automating this process enhances the efficiency and accuracy of clinical records processing. However, current methods struggle with noisy and lengthy clinical texts, making it difficult to ensure the reliability of feature extraction. Furthermore, they typically make separate binary predictions for each code, overlooking the dependencies between them. To address these issues, we propose a novel model called Hierarchical Feature Optimization and Rescorer (HiRes). We employ a cascaded convolution architecture to mitigate noise and enhance feature representation. Additionally, a masked autoencoder-based rescorer is introduced to capture ICD code interdependencies, refining initial predictions for improved accuracy. The model also considers the inconsistencies in code representations. Experiments on the MIMIC datasets demonstrate the effectiveness of our model. © 2025 IEEE.

关键词： automatic ICD coding cascaded convolutional network multi-label classification

来源：评论

学校读者我要写书评

暂无评论

ALFC-Point: Adaptive Laplacian Feature Convolution network for 3D Point Cloud Understanding 41st

ALFC-Point: Adaptive Laplacian Feature Convolution Network f...

引用

41st computer Graphics International Conference, CGI 2024

作者： Liao, Junjie Yin, Mengxiao Li, Ming Zhu, Congyang Yang, Zhiqiang Zhan, Feng School of Computer and Electronics Information Guangxi University Guangxi Nanning530004 China Guangxi Key Laboratory of Multimedia Communications and Network Technology Guangxi Nanning530004 China Education Department of Guangxi Zhuang Autonomous Region Nanning China

ISBN: (纸本)9783031820236

To capture 3D geometric details, point cloud understanding primarily employs convolution, dynamic map, or transformer methods to build complex local geometry extractors. When dealing with sparse and noisy point clouds in real-world environments, the structural design of these methods tends to be complex and perform poorly. In this paper, we propose a lightweight point cloud understanding framework based on Laplacian feature convolution, which has high accuracy on normal-sized point cloud datasets and performs well when dealing with sparse and noisy point cloud inputs. The adaptive Laplacian convolution module combined with the vector attention mechanism can effectively capture the local details of point cloud geometry. The framework employs a multi-scale feature fusion mechanism between modules to reduce the number of sampling points step by step, reducing redundant computations and achieving a lightweight effect. In addition, we propose a parameter-free module for interpolation to enhance the training part. Experimental results on object-level datasets show that the lightweight method proposed in our paper has similar accuracy to the current best method in terms of accuracy (Acc) and mean intersection over union (mIoU), and has better robustness in handling sparse and noisy point clouds. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Laplace transforms

来源：评论

学校读者我要写书评

暂无评论

Multi-routine-data driven spatio-temporal short-term predictions for surface ozone in China

引用

Air Quality, Atmosphere and Health 2025年 1-10页

作者： Zheng, Canju Shen, Hengqing Sun, Jianan Liu, Guangliang Cao, Haowei Zhang, Jie Gong, Xiang Xu, Da Key Laboratory of Computing Power Network and Information Security Ministry of Education Shandong Computer Science Center Qilu University of Technology (Shandong Academy of Sciences) Jinan China Environment Research Institute Shandong University Shandong Qingdao 266237 China Ningbo Yonghuanyuan Environmental Engineering and Technology Co. LTD Zhejiang Ningbo 315012 China School of Mathematics and Physics Qingdao University of Science and Technology Qingdao 266061 China Shandong Provincial Key Laboratory of Computer Networks Shandong Fundamental Research Center for Computer Science Jinan China Qingdao Innovation Center of Artificial Intelligence Ocean Technology Qingdao 266061 China

Ozone (O3) is a major atmospheric pollutant, and accurate prediction of its concentrations remains challenging due to its complex nonlinear relationships with precursor compounds. Existing machine learning methods hav... 详细信息

Ozone (O₃) is a major atmospheric pollutant, and accurate prediction of its concentrations remains challenging due to its complex nonlinear relationships with precursor compounds. Existing machine learning methods have mainly focused on single-site or spatial predictions, lacking research on spatio-temporal short-term predictions based on simple factors. To address this gap, the MRD-O₃former, a deep learning model driven by multi-routine data, was developed to predict short-term hourly spatio–temporal surface ozone concentrations over China. The model exhibits strong spatio–temporal consistency, achieving a high correlation coefficient (r2=0.85∼0.90) and low normalized mean biases (NMBs) between -15% and 15% at the national scale compared to reanalysis ozone data. Both ablation experiments and permutation importance analysis reveal that historical ozone levels play a primary role in next-day ozone prediction, while meteorological factors such as wind speed and temperature also make significant contributions. Regional validation confirms the model’s effectiveness in the Beijing-Tianjin-Hebei region. Moreover, the study investigates the differential impact of crucial factors in urban and rural areas, revealing that historical ozone levels and meteorological factors significantly influence rural areas. However, the influence of historical ozone levels on urban ozone prediction is relatively small, especially during the summer. This suggests that urban ozone undergoes rapid formation and removal processes. These findings highlight the promising potential of deep learning techniques in accurately predicting spatiotemporal short-term ozone concentrations and interpreting the mechanism and source for ozone pollution. © The Author(s), under exclusive licence to Springer Nature B.V. 2025.

关键词： Data-driven Multi-factor Ozone concentration Spatio-temporal prediction

来源：评论

学校读者我要写书评

暂无评论

CroPrompt: Cross-task Interactive Prompting for Zero-shot Spoken Language Understanding

CroPrompt: Cross-task Interactive Prompting for Zero-shot Sp...

引用

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

作者： Qin, Libo Wei, Fuxuan Chen, Qiguang Zhou, Jingxuan Huang, Shijue Si, Jiasheng Lu, Wenpeng Che, Wanxiang School of Computer Science and Engineering Central South University China Key Laboratory of Data Intelligence and Advanced Computing in Provincial Universities Soochow University China Research Center for Social Computing and Information Retrieval Harbin Institute of Technology China Harbin Institute of Technology Shenzhen China Key Laboratory of Computing Power Network and Information Security Ministry of Education Qilu University of Technology Shandong Academy of Sciences China

ISBN: (纸本)9798350368741

Slot filling and intent detection are two highly correlated tasks in spoken language understanding (SLU). Recent SLU research attempts to explore zero-shot prompting techniques in large language models to alleviate the data scarcity problem. Nevertheless, the existing prompting work ignores the cross-task interaction information for SLU, which leads to sub-optimal performance. To solve this problem, we present the pioneering work of Cross-task Interactive Prompting (CroPrompt) for SLU, which enables the model to interactively leverage the information exchange across the correlated tasks in SLU. Additionally, we further introduce a multi-task self-consistency mechanism to mitigate the error propagation caused by the intent information injection. We conduct extensive experiments on the standard SLU benchmark and the results reveal that CroPrompt consistently outperforms the existing prompting approaches. In addition, the multi-task self-consistency mechanism can effectively ease the error propagation issue, thereby enhancing the performance. We hope this work can inspire more research on cross-task prompting for SLU. © 2025 IEEE.

关键词：

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：