检索结果-南通市图书馆

Robust video question answering via contrastive cross-modality representation learning

在线全文

学校读者我要写书评

暂无评论

Science China(information Sciences) 2024年第10期67卷 211-226页

作者： Xun YANG Jianming ZENG Dan GUO Shanshan WANG Jianfeng DONG Meng WANG School of information Science and Technology University of Science and Technology of China Institute of Artificial Intelligence Hefei Comprehensive National Science Center School of Computer Science and information Engineering Hefei University of Technology Institutes of Physical Science and information Technology Anhui University School of Computer Science and Technology Zhejiang Gongshang University

Video question answering(VideoQA) is a challenging yet important task that requires a joint understanding of low-level video content and high-level textual semantics. Despite the promising progress of existing efforts, recent studies revealed that current VideoQA models mostly tend to over-rely on the superficial correlations rooted in the dataset bias while overlooking the key video content, thus leading to unreliable results. Effectively understanding and modeling the temporal and semantic characteristics of a given video for robust VideoQA is crucial but, to our knowledge, has not been well investigated. To fill the research gap, we propose a robust VideoQA framework that can effectively model the cross-modality fusion and enforce the model to focus on the temporal and global content of videos when making a QA decision instead of exploiting the shortcuts in datasets. Specifically, we design a self-supervised contrastive learning objective to contrast the positive and negative pairs of multimodal input, where the fused representation of the original multimodal input is enforced to be closer to that of the intervened input based on video perturbation. We expect the fused representation to focus more on the global context of videos rather than some static keyframes. Moreover, we introduce an effective temporal order regularization to enforce the inherent sequential structure of videos for video representation. We also design a Kullback-Leibler divergence-based perturbation invariance regularization of the predicted answer distribution to improve the robustness of the model against temporal content perturbation of videos. Our method is model-agnostic and can be easily compatible with various VideoQA backbones. Extensive experimental results and analyses on several public datasets show the advantage of our method over the state-of-the-art methods in terms of both accuracy and robustness.

关键词： video question answering cross-modality fusion contrastive learning cross-media reasoning

An angular blinking jamming method based on electronically controlled corner reflectors

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Journal of Systems Engineering and Electronics 2024年第2期35卷 330-338页

作者： GAN Lin WU Zehao WANG Xuesong LI Jianbing state Key Laboratory of Complex Electromagnetic Environmental Effects on Electronics and information System National University of Defense TechnologyChangsha 410073China

Passive jamming is believed to have very good potential in countermeasure *** this paper,a passive angular blinking jamming method based on electronically controlled corner reflectors is *** amplitude of the incident wave can be modulated by switching the corner reflector between the penetration state and the reflection state,and the ensemble of multiple corner reflectors with towing rope can result in complex angle decoying *** of the decoying effect on corner reflectors’radar cross section and positions are analyzed and *** show that the angle measured by a monopulse radar can be significantly interfered by this method while the automatic tracking is employed.

关键词： monopulse radar angular blinking jamming corner reflector amplitude-comparison angle measurement metasurface

Resource Allocation for Channel Estimation in Reconfigurable Intelligent Surface-Aided Multi-Cell Networks

在线全文

学校读者我要写书评

暂无评论

Journal of Communications and information Networks 2024年第1期9卷 64-79页

作者： Yining Xu Sheng Zhou Department of Electronic Engineering Beijing National Research Center for Information Science and TechnologyTsinghua UniversityBeijing 100084China

Reconfigurable intelligent surface(RIS)is a promising solution to deal with the blockage-sensitivity of millimeter wave band and reduce the high energy consumption caused by network densification. However, deploying large scale RISs may not bring expected performance gain due to significant channel estimation overhead and non-negligible reflected *** this paper,we derive the analytical expressions of the coverage probability, area spectrum efficiency(ASE)and energy efficiency (EE)of a downlink RIS-aided multi-cell *** order to optimize the network performance, we investigate the conditions for the optimal number of training symbols of each antenna-to-antenna and antenna-to-element path (referred to as the optimal unit training overhead) in channel *** study shows that:1)RIS deployment is not“the more, the better”, only when blockage objects are dense should one deploy more RISs;2) the coverage probability is maximized when the unit training overhead is designed as large as possible;3)however,the ASE-and-EE-optimal unit training overhead exists. It is a monotonically increasing function of the frame length and a monotonically decreasing function of the average signal-to-noise-ratio (in the high signal-to-noise-ratio region). Additionally,the optimal unit training overhead is smaller when communication nodes deploy particularly few or many antennas.

关键词： reconfigurable intelligent surface multi-cell network directional transmissions channel estimation resource allocation stochastic geometry

Predicting Users’ Latent Suicidal Risk in Social Media: An Ensemble Model Based on Social Network Relationships

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Computers, Materials & Continua 2024年第6期79卷 4259-4281页

作者： Xiuyang Meng Chunling Wang Jingran Yang Mairui Li Yue Zhang Luo Wang School of information Science and Technology Beijing Forestry UniversityBeijing100083China Engineering Research center for Forestry-Oriented Intelligent information Processing of National Forestry and Grassland Administration Beijing100083China

Suicide has become a critical concern,necessitating the development of effective preventative *** media platforms offer a valuable resource for identifying signs of suicidal *** progress in detecting suicidal ideation on social media,accurately identifying individuals who express suicidal thoughts less openly or infrequently poses a significant *** tackle this,we have developed a dataset focused on Chinese suicide narratives from Weibo’s Tree Hole feature and introduced an ensemble model named Text Convolutional Neural Network based on Social Network relationships(TCNN-SN).This model enhances predictive performance by leveraging social network relationship features and applying correction factors within a weighted linear fusion *** is specifically designed to identify key individuals who can help uncover hidden suicidal users and *** model,assessed using the bespoke dataset and benchmarked against alternative classification approaches,demonstrates superior accuracy,F1-score and AUC metrics,achieving 88.57%,88.75%and 94.25%,respectively,outperforming traditional TextCNN models by 12.18%,10.84%and 10.85%.We assert that our methodology offers a significant advancement in the predictive identification of individuals at risk,thereby contributing to the prevention and reduction of suicide incidences.

关键词： Suicide risk prediction social media social network relationships Weibo Tree Hole deep learning

Understanding adversarial attacks on observations in deep reinforcement learning

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Science China(information Sciences) 2024年第5期67卷 69-83页

作者： You QIAOBEN Chengyang YING Xinning ZHOU Hang SU Jun ZHU Bo ZHANG Department of Computer Science and Technology Beijing National Research Center for Information Science and Technology Tsinghua-Bosch Joint Center for Machine Learning Institute for Artificial Intelligence Tsinghua University Peng Cheng Laboratory

Deep reinforcement learning models are vulnerable to adversarial attacks that can decrease the cumulative expected reward of a victim by manipulating its observations. Despite the efficiency of previous optimization-based methods for generating adversarial noise in supervised learning, such methods might not achieve the lowest cumulative reward since they do not generally explore the environmental ***, a framework is provided to better understand the existing methods by reformulating the problem of adversarial attacks on reinforcement learning in the function space. The reformulation approach adopted herein generates an optimal adversary in the function space of targeted attacks, repelling them via a generic two-stage framework. In the first stage, a deceptive policy is trained by hacking the environment and discovering a set of trajectories routing to the lowest reward or the worst-case performance. Next, the adversary misleads the victim to imitate the deceptive policy by perturbing the observations. Compared to existing approaches, it is theoretically shown that our adversary is strong under an appropriate noise level. Extensive experiments demonstrate the superiority of the proposed method in terms of efficiency and effectiveness, achieving state-of-the-art performance in both Atari and MuJoCo environments.

关键词： deep learning reinforcement learning adversarial robustness adversarial attack

Efficient Penetration Testing Path Planning Based on Reinforcement Learning with Episodic Memory

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Computer Modeling in Engineering & Sciences 2024年第9期140卷 2613-2634页

作者： Ziqiao Zhou Tianyang Zhou Jinghao Xu Junhu Zhu Henan Key Laboratory of information Security National Engineering Technology Research Center of the Digital Switching SystemZhengzhou450000China School of Cryptographic Engineering Information Engineering UniversityZhengzhou450000China

Intelligent penetration testing is of great significance for the improvement of the security of information systems,and the critical issue is the planning of penetration test *** view of the difficulty for attackers to obtain complete network information in realistic network scenarios,Reinforcement Learning(RL)is a promising solution to discover the optimal penetration path under incomplete information about the target *** RL-based methods are challenged by the sizeable discrete action space,which leads to difficulties in the ***,most methods still rely on experts’*** address these issues,this paper proposes a penetration path planning method based on reinforcement learning with episodic ***,the penetration testing problem is formally described in terms of reinforcement *** speed up the training process without specific prior knowledge,the proposed algorithm introduces episodic memory to store experienced advantageous strategies for the first ***,the method offers an exploration strategy based on episodic memory to guide the agents in *** design makes full use of historical experience to achieve the purpose of reducing blind exploration and improving planning ***,comparison experiments are carried out with the existing RL-based *** results reveal that the proposed method has better convergence *** running time is reduced by more than 20%.

关键词： Intelligent penetration testing penetration testing path planning reinforcement learning episodic memory exploration strategy

Multi-Branch Fault Line Location Method Based on Time Difference Matrix Fitting

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Energy Engineering 2024年第1期121卷 77-94页

作者： Hua Leng Silin He Jian Qiu Feng Liu Xinfei Huang Jiran Zhu School of Electrical and information Engineering Hunan UniversityChangsha410082China Distribution Network Technical center State Grid Hunan Electric Power Research InstituteChangsha410007China Technical center China Energy Construction Group Hunan Thermal Power ConstructionZhuzhou412011China School of Electrical and information Engineering Changsha University of Science and TechnologyChangsha410076China

The distribution network exhibits complex structural characteristics,which makes fault localization a challenging *** when a branch of the multi-branch distribution network fails,the traditional multi-branch fault location algorithm makes it difficult to meet the demands of high-precision fault localization in the multi-branch distribution network *** this paper,the multi-branch mainline is decomposed into single branch lines,transforming the complex multi-branch fault location problem into a double-ended fault location *** on the different transmission characteristics of the fault-traveling wave in fault lines and non-fault lines,the endpoint reference time difference matrix S and the fault time difference matrix G were *** time variation rule of the fault-traveling wave arriving at each endpoint before and after a fault was comprehensively *** realize the fault segment location,the least square method was *** was used to find the first-order fitting relation that satisfies the matching relationship between the corresponding row vector and the first-order function in the two matrices,to realize the fault segment ***,the time difference matrix is used to determine the traveling wave velocity,which,combined with the double-ended traveling wave location,enables accurate fault location.

关键词： Multi-branch lines distribution network fault location double-ended traveling wave positioning least square method

In-situ deposited anti-aging TiN capping layer for Nb superconducting quantum circuits

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Chinese Physics B 2024年第9期33卷 133-137页

作者： Hao-Ran Tao Lei Du Liang-Liang Guo Yong Chen Hai-Feng Zhang Xiao-Yan Yang Guo-Liang Xu Chi Zhang Zhi-Long Jia Peng Duan Guo-Ping Guo CAS Key Laboratory of Quantum information University of Science and Technology of ChinaHefei 230026China CAS center for Excellence and Synergetic Innovation center in Quantum information and Quantum Physics University of Science and Technology of ChinaHefei 230026China Origin Quantum Computing Company Limited Hefei 230088China

The performance of Nb superconducting quantum devices is predominantly limited by dielectric loss at the metal–air interface,where Nb2O5 is considered the main loss ***,we suppress the formation of native oxides by in-situ deposition of a TiN capping layer on the Nb *** TiN capping layers,no Nb2O5 forms on the surface of the Nb *** quality factor Qi of the Nb resonator increases from 5.6×10^(5) to 7.9×10^(5) at low input power and from 6.8×10^(6) to 1.1×10^(7)at high input ***,the TiN capping layer also shows good aging resistance in Nb resonator devices,with no significant performance fluctuations after one month of *** findings highlight the effectiveness of TiN capping layers in enhancing the performance and longevity of Nb superconducting quantum devices.

关键词： anti-aging oxidation dielectric loss Nb superconducting quantum circuits

GLOBAL CONVERGENCE OF A CAUTIOUS PROJECTION BFGS ALGORITHM FOR NONCONVEX PROBLEMS WITHOUT GRADIENT LIPSCHITZ CONTINUITY

在线全文

学校读者我要写书评

暂无评论

Acta Mathematica Scientia 2024年第5期44卷 1735-1746页

作者： Gonglin YUAN Xiong ZHAO Jiajia YU School of Mathematics and information Science Center for Applied Mathematics of Guangxi(Guangxi University)Guangxi UniversityNanning530004China

A cautious projection BFGS method is proposed for solving nonconvex unconstrained optimization *** global convergence of this method as well as a stronger general convergence result can be proven without a gradient Lipschitz continuity assumption,which is more in line with the actual problems than the existing modified BFGS methods and the traditional BFGS *** some additional conditions,the method presented has a superlinear convergence rate,which can be regarded as an extension and supplement of BFGS-type methods with the projection ***,the effectiveness and application prospects of the proposed method are verified by numerical experiments.

关键词： cautious BFGS nonconvex problems Lipschitz continuity projection technique global convergence