检索结果-内蒙古大学图书馆

arXiv 2024年

作者： Ma, Yingjie Yu, Zitong Lin, Xun Xie, Weicheng Shen, Linlin Great Bay University China College of Computer Science and Software Engineering Shenzhen University China National Engineering Laboratory for Big Data System Computing Technology Shenzhen University China Guangdong Provincial Key Laboratory of Intelligent Information Processing China

In the domain of facial recognition security, multimodal Face Anti-Spoofing (FAS) is essential for countering presentation attacks. However, existing technologies encounter challenges due to modality biases and imbalances, as well as domain shifts. Our research introduces a Mixture of Experts (MoE) model to address these issues effectively. We identified three limitations in traditional MoE approaches to multimodal FAS: (1) Coarse-grained experts’ inability to capture nuanced spoofing indicators;(2) Gated networks’ susceptibility to input noise affecting decision-making;(3) MoE’s sensitivity to prompt tokens leading to overfitting with conventional learning methods. To mitigate these, we propose the Bypass Isolated Gating MoE (BIG-MoE) framework, featuring: (1) Fine-grained experts for enhanced detection of subtle spoofing cues;(2) An isolation gating mechanism to counteract input noise;(3) A novel differential convolutional prompt bypass enriching the gating network with critical local features, thereby improving perceptual capabilities. Extensive experiments on four benchmark datasets demonstrate significant generalization performance improvement in multimodal FAS task. The code is released at https://***/murInJ/BIGMoE. Copyright © 2024, The Authors. All rights reserved.

关键词： HTTP

来源：评论

学校读者我要写书评

暂无评论

Impact Investigation of Adversarial Samples on CNN-Based SAR Image Target Detectors

Impact Investigation of Adversarial Samples on CNN-Based SAR...

引用

Pattern Recognition and Artificial Intelligence (PRAI), International Conference on

作者： Peng Zhou Shunping Xiao Mingdian Li Junwu Deng Xiangwei Xing Siwei Chen College of Electronic Science and Technology National University of Defense Technology Changsha China Satellite Information Intelligent Processing and Application Research Laboratory Beijing China

In recent years, deep learning algorithms based on convolutional neural networks (CNN) have made breakthroughs in the field of target detection in Synthetic Aperture Radar (SAR) images. However, CNN may be vulnerable to adversarial examples, raising security concerns for practical deployment. To gain a deeper understanding of adversarial examples and provide a theoretical basis for building robust detection networks, this study comprehensively investigates the impacts of adversarial samples generated by deep recognition networks (DRN) in SAR images on the performance of target detection networks with different principles, using YOLOv3 and Faster-RCNN as examples. Experimental results on the MiniSAR vehicle target dataset show that adversarial samples pose a more serious threat to the single-stage detection network YOLOv3. Simply adding small perturbation that are invisible for human eyes to the SAR image can mislead the detector into missing the majority of targets. In contrast, two-stage detection networks like Faster-RCNN exhibit stronger robustness with their detection performance being less influenced by the adversarial examples generated by different DRN.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Explicit Predictive Control of Forward DC-DC Converter Based on an Input-Mapping Method

Explicit Predictive Control of Forward DC-DC Converter Based...

引用

第43届中国控制会议

作者： Xiang Gu Dewei Li Aoyun Ma Yaru Yu Department of Automation Shanghai Jiao Tong University Key Laboratory of System Control and Information Processing Ministry of Education of China Shanghai Engineering Research Center of Intelligent Control and Management

ISBN: (数字)9789887581581

ISBN: (纸本)9798350366907

In the practical application of model predictive control in power electronics, both online computational burden and model accuracy are crucial. This paper presents an explicit model predictive control method based on input-mapping *** address the issue of parameter drift in the forward DC-DC converter system, the input-mapping method is introduced to compensate for model deviations using historical data. Firstly, the average model of the forward DC-DC converter system is introduced. For the control objective of voltage tracking, an explicit predictive control using the incremental model is ***, input-mapping method is integrated with explicit predictive control. During offline computation, analytical forms of the input-mapping combination coefficients and the optimal solution function of explicit model predictive control are *** computation of combination coefficients is conducted to obtain the optimal solution rapidly. The effectiveness of the proposed method is verified through simulations in MATLAB.

关键词： Forward DC-DC converter explicit model predictive control input-mapping method

来源：评论

学校读者我要写书评

暂无评论

TRAIL: Trust-Aware Client Scheduling for Semi-Decentralized Federated Learning

arXiv

引用

arXiv 2024年

作者： Hu, Gangqiang Lu, Jianfeng Han, Jianmin Cao, Shuqin Liu, Jing Fu, Hao School of Computer Science and Technology Zhejiang Normal University China School of Computer Science and Technology Wuhan University of Science and Technology China Key Laboratory of Social Computing and Cognitive Intelligence Dalian University of Technology Ministry of Education China Hubei Province Key Laboratory of Intelligent Information Processing and Real-time Industrial System Wuhan Universityof Science and Technology China

Due to the sensitivity of data, Federated Learning (FL) is employed to enable distributed machine learning while safeguarding data privacy and accommodating the requirements of various devices. However, in the context of semidecentralized FL, clients' communication and training states are dynamic. This variability arises from local training fluctuations, heterogeneous data distributions, and intermittent client participation. Most existing studies primarily focus on stable client states, neglecting the dynamic challenges inherent in real-world scenarios. To tackle this issue, we propose a TRust-Aware clIent scheduLing mechanism called TRAIL, which assesses client states and contributions, enhancing model training efficiency through selective client participation. We focus on a semi-decentralized FL framework where edge servers and clients train a shared global model using unreliable intra-cluster model aggregation and inter-cluster model consensus. First, we propose an adaptive hidden semi- Markov model to estimate clients' communication states and contributions. Next, we address a client-server association optimization problem to minimize global training loss. Using convergence analysis, we propose a greedy client scheduling algorithm. Finally, our experiments conducted on real-world datasets demonstrate that TRAIL outperforms state-of-the-art baselines, achieving an improvement of 8.7% in test accuracy and a reduction of 15.3% in training loss. © 2024, CC BY.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Enhancing Drug-Drug Interaction Prediction by Knowledge Graph Embedding

Enhancing Drug-Drug Interaction Prediction by Knowledge Grap...

引用

International Conference on Machine Learning and Cybernetics (ICMLC)

作者： CHAO LIU XIZHAO WANG QIANG LIU Big Data Institute College of Computer Science and Software Engineering Shenzhen University Shenzhen China Guangdong Key Laboratory of Intelligent Information Processing Shenzhen University Shenzhen China

ISBN: (数字)9798331528041

ISBN: (纸本)9798331528058

Drug-drug interactions (DDIs) refer to the synergistic or antagonistic effects between different drugs. Synergistic effects can enhance therapeutic efficacy, while antagonistic effects may reduce efficacy or even trigger adverse reactions, worsening the patient's condition. Most existing DDI prediction studies overlook the features of chemical bonds between atoms within drug molecules and the interaction types between drug pairs, which to some extent limits the accuracy and reliability of the predictions. In view of this, we have designed a novel framework called KGE-DDI, which leverages knowledge graph embedding (KGE) technology to enhance the prediction of DDIs. Specifically, for individual drug molecules, KGE-DDI first integrates the features of atoms and chemical bond. Then, by introducing attention mechanisms and graph neural network technology, it learns representations of the drug molecules. For pairs of drug molecules, KGE-DDI models their interactions utilizing atom-level Pearson correlation matrices. Finally, it predicts the interactions between drug pairs by employing our designed scoring function. We conducted comprehensive experiments under both transductive and inductive settings, and the results demonstrate the effectiveness and superiority of KGE-DDI in three scenarios: (existing drug, existing drug), (new drug, new drug), and (new drug, existing drug).

关键词： Drugs Measurement Correlation Knowledge graphs Atoms Predictive models Graph neural networks Reliability Chemicals Cybernetics

来源：评论

学校读者我要写书评

暂无评论

Performance Bounds and Optimization for CSI-Ratio based Bi-static Doppler Sensing in ISAC Systems

arXiv

引用

arXiv 2024年

作者： Hu, Yanmo Wu, Kai Andrew Zhang, J. Deng, Weibo Jay Guo, Y. Key Laboratory of Marine Environmental Monitoring and Information Processing Ministry of Industry and Information Technology China School of Electronic and Information Engineering Harbin Institute of Technology China Global Big Data Technologies Centre The University of Technology Sydney Australia

Bi-static sensing is crucial for exploring the potential of networked sensing capabilities in integrated sensing and communications (ISAC). However, it suffers from the challenging clock asynchronism issue. CSI ratio-based sensing is an effective means to address the issue. Its performance bounds, particular for Doppler sensing, have not been fully understood yet. This work endeavors to fill the research gap. Focusing on a single dynamic path in high-SNR scenarios, we derive the closed-form CRB. Then, through analyzing the mutual interference between dynamic and static paths, we simplify the CRB results by deriving close approximations, further unveiling new insights of the impact of numerous physical parameters on Doppler sensing. Moreover, utilizing the new CRB and analyses, we propose novel waveform optimization strategies for noise- and interference-limited sensing scenarios, which are also empowered by closed-form and efficient solutions. Extensive simulation results are provided to validate the preciseness of the derived CRB results and analyses, with the aid of the maximum-likelihood estimator. The results also demonstrate the substantial enhanced Doppler sensing accuracy and the sensing capabilities for low-speed target achieved by the proposed waveform design. © 2024, CC BY.

关键词： Signal to noise ratio

来源：评论

学校读者我要写书评

暂无评论

Milliwatt-threshold widely-tunable optical parametric oscillation in integrated chalcogenide microresonators

Milliwatt-threshold widely-tunable optical parametric oscill...

引用

2022 Conference on Lasers and Electro-Optics Pacific Rim, CLEO/PR 2022

作者： Wang, Zifu Zhao, Jiaxin Xia, Di Li, Yufei Luo, Liyang Liu, Dong Zhang, Bin Li, Zhaohui Guangdong Provincial Key Laboratory of Optoelectronic Information Processing Chips and Systems School of Electrical and Information Technology Sun Yat-sen University Guangzhou510275 China Key Laboratory of Optoelectronic Materials and Technologies Sun Yat-sen University Guangzhou510275 China Zhuhai519000 China

关键词： Microresonators

来源：评论

学校读者我要写书评

暂无评论

iWavePro: An improved framework for iWave++

iWavePro: An improved framework for iWave++

引用

IEEE International Symposium on Circuits and Systems (ISCAS)

作者： Dongmei Xue Cunhui Dong Fan Ye Hang Chen Bowei Kang Li Li Dong Liu CAS Key Laboratory of Technology in Geo-Spatial Information Processing and Application System University of Science and Technology of China Hefei China

End-to-end image coding methods based on wavelet-like transform have made great progress in recent years. The most advanced one is iWave++, which adopts multi-level lifting schemes based on convolutional neural networks. However, iWave++ still has many unresolved problems. First, the independent entropy coding of each component makes it impossible to use the correlation between components better. Secondly, additive wavelet transform limits the nonlinear ability of learnable wavelet transform. Moreover, the offline training strategy makes the iWave++ unable to adjust according to the content. In this paper, we propose an improved framework for iWave++ called iWave-Pro. iWavePro is designed with several techniques to overcome the problems mentioned above. These techniques are the joint multi-component Gaussian mixture entropy coding, the affine wavelet-like transform, and the online training. Experimental results show that our method can save 10.73% bit rate compared with iWave++ at the same quality.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Power Allocation Based on CNN and Long Short-Term Memory Networks in Massive MIMO Systems 12th

Power Allocation Based on CNN and Long Short-Term Memory Net...

引用

12th International Conference on Computer Engineering and Networks, CENet 2022

作者： Cui, Zijian Tan, Fangqing School of Information and Communication Guilin University of Electronic Technology Guilin541004 China Guangxi Key Laboratory of Wireless Wideband Communication and Signal Processing Guilin University of Electronic Technology Guilin541004 China

ISBN: (纸本)9789811969003

Massive MIMO is one of the key technologies in 5G and 6G wireless networks, with advantages of low complexity, high throughput, ultra-reliability and low latency. In this study, convolutional neural networks (CNN) and long short-term memory networks are used to carry out power allocation strategies in the downlink of Massive MIMO networks. In particular, deep neural networks are used to learn mathematical relationships between location and the power allocation strategies of user devices. And this trained mapping is used to predict the power allocation of new user devices. Deep learning can significantly reduce the computational complexity of power allocation compared to traditional methods. However, the structure of the existing neural network based on particular scene is relatively simple and basic. Considering the rapid progress of algorithms in the field of recent deep learning, there is a large room for improvement. To this end, the accuracy and performance of the models are improved using CNN and long short-term memory network. Simulation results show that two optimized network models can improve the loss to a lower level in shorter time than fully connected network. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： MIMO systems

来源：评论

学校读者我要写书评

暂无评论

Knowledge tracking model based on recurrent neural network and transformer 8

Knowledge tracking model based on recurrent neural network a...

引用

8th International Symposium on Advances in Electrical, Electronics, and Computer Engineering, ISAEECE 2023

作者： Cheng, Yan Zhao, Songhua Hu, Jiansheng Zou, Haifeng Luo, Pin Fu, Yan Zhong, Linhui Liu, Chunlei School of Computer and Information Engineering Jiangxi Normal University Nanchang330022 China Jiangxi Province Science and Technology Innovation Base-Provincial Key Laboratory "Jiangxi Province Key Laboratory of Intelligent Education" Nangchang330022 China Jiangxi Ruanyun Technology Corporation Limited Nanchang330096 China Jiangxi Heyi Keji Youxian Gongsi Nanchang330096 China

ISBN: (纸本)9781510666290

With the continuous development of online education platform, knowledge tracking (KT) has become a key technology to help online education platform provide personalized education. However, the existing knowledge tracking model based on recurrent neural network is difficult to be used for the input of long sequence, and has the problem of long-term dependence. Secondly, although the knowledge tracking model based on Transformer does not have the problem of long-term dependence, it is difficult to capture the input sequence information. Therefore, this paper proposes a knowledge tracking model based on recurrent neural network and transformer. A new position coding method is designed, and LSTM is used to replace the position coding method of Transformer to encode sequence features, so that the model in this paper can not only capture the input sequence information, but also get rid of the long-term dependency problem based on the recurrent neural network, and use GRU network to capture the context information. In addition, an adaptive fusion gate is designed to fuse the global features and context features obtained by Transformer, and use the fused features to predict the students' answers to the next question. In addition, an adaptive fusion gate is designed to fuse the global features and context features obtained by Transformer, and use the fused features to predict the students' answers to the next question. © 2023 SPIE. All rights reserved.

关键词： Students

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：