检索结果-南通市图书馆

Robust video question answering via contrastive cross-modality representation learning

维普期刊数据库评论

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

Physics-informed deep Koopman operator for Lagrangian dynamic systems

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

science China(Information sciences) 2024年第9期67卷 146-163页

作者： Xuefeng WANG Yang CAO Shaofeng CHEN Yu KANG Department of Automation University of Science and Technology of China State Key Laboratory of Fire science University of Science and Technology of China Institute of Advanced technology University of Science and Technology of China

Accurate mechanical system models are crucial for safe and stable control. Unlike linear systems,Lagrangian systems are highly nonlinear and difficult to optimize because of their unknown system *** research thus used deep neural networks to generate linear models of original systems by mapping nonlinear dynamic systems into a linear space with a Koopman observable function encoder. The controller then relies on the Koopman linear model. However, without physical information constraints, ensuring control consistency between the original nonlinear system and the Koopman system is tough, as the learning process of the Koopman observation function is unsupervised. This paper thus proposes a two-stage learning algorithm that uses structural subnetworks to build a physics-informed network topology to simultaneously learn the Koopman observable functions and the system energy representation. In the Koopman matrix learning session, a quadratic-constrained optimization problem is solved to ensure that the Koopman representation satisfies the energy difference matching hard constraint. The proposed energy-preserving deep Lagrangian Koopman(EPDLK) framework effectively represents the dynamics of the Lagrangian system while ensuring control consistency. The effectiveness of EPDLK is compared with those of various Koopman observable function construction methods in multistep prediction and trajectory tracking tasks. EPDLK achieves better control consistency by guaranteeing energy difference matching, which facilitates the application of the control law generated on the Koopman system directly to the original nonlinear Lagrangian system.

关键词： Koopman operator Lagrangian physics-informed nonlinear model predictive control (MPC)

Carbon felt electrode coated with WS2 enables a high-performance polysulfide/ferricyanide flow battery

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Rare Metals 2024年第10期43卷 5039-5047页

作者： Liang-Yu Li Su Yan Ying-Jia Huang Fang-Fang Zhong Jin-Chao Cao Mei Ding Chuan-Kun Jia College of Materials science and Engineering Changsha University of Science and Technology Institute of Energy Storage technology College of Energy and Power EngineeringChangsha University of Science and Technology Binzhou Institute of technology Weiqiao-UCAS Science and Technology Park

Polysulfide/ferricyanide flow batteries （S/Fe RFBs）,with the advantages of abundant earth reservation low cost,high safety,and environmental friendliness,have attracted significant interest and demonstrated noteworthy potential for practical ***,the battery performance,including the energy efficiency （EE）,voltage efficiency （VE）,and power density of the S/Fe RFBs remains low owing to the slow redox kinetics of polysulfide *** address these concerns,WS2was selected as the booster and deposited on a commercial carbon felt electrode （WS2–CF） to stimulate the redox reactions of polysulfide *** better hydrophilicity and smaller charge-transfer resistance,WS2–CF exhibits enhanced electrochemical activity toward polysulfide redox ***,the battery performance of S/Fe RFB with WS2–CF as the anode has been improved,with EE of 84%,VE of 84%,and a peak power density of 175.7 mW·cm-2,which are all higher than the cell only with the bare carbon felt （CF） as electrodes (76%,77% and 155.8 mW·cm-2,respectively).Furthermore,the cycling life of the S/Fe RFB with WS2–CF has been prolonged to 2200 cycles with a capacity retention of 96% a 40 mA·cm-2because of the good stability of WS2–CF as the ***,under the same conditions,the S/Fe RFB without WS2–CF terminated after 1500 cycles with a fast capacity *** successful utilization of WS2as a booster on an electrode provides an efficient strategy for obtaining advanced S/Fe RFBs for practical applications.

关键词： Carbon felt Energy storage Polysulfide/ferricyanide flow battery Tungsten disulfide

Effect of strain on structure and electronic properties of monolayer C_(4)N_(4)

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Chinese Physics B 2024年第5期33卷 595-600页

作者：陈昊徐瑛赵家石周丹 School of Physics Changchun University of Science and TechnologyChangchun 130022China School of Computer science and technology Changchun University of Science and TechnologyChangchun 130022China

The first-principles calculations are performed to examine structural,mechanical,and electronic properties at large strain for a monolayer C_(4)N_(4),which has been predicted as an anchoring promising material to attenuate shuttle effect in Li–S batteries stemming from its large absorption energy and low diffusion energy *** results show that the ideal strengths of C_(4)N_(4)under tension and pure shear deformation conditions reach 13.9 GPa and 12.5 GPa when the strains are 0.07 and 0.28,*** folded five-membered rings and diverse bonding modes between carbon and nitrogen atoms enhance the ability to resist plastic deformation of C_(4)N_(4).The orderly bond-rearranging behaviors under the weak tensile loading path along the[100]direction cause the impressive semiconductor–metal transition and inverse semiconductor–metal *** present results enrich the knowledge of the structure and electronic properties of C_(4)N_(4)under deformations and shed light on exploring other two-dimensional materials under diverse loading conditions.

关键词： two-dimensional materials strain effect structural evolution electronic properties

维普期刊数据库

Triphenylphosphine-Assisted Exsolution Engineering on Ruddlesden-Popper Perovskites for Promoting Oxygen Evolution

在线全文

学校读者我要写书评

暂无评论

Energy & Environmental Materials 2024年第4期7卷 433-441页

作者： Juan Bai Jing Shang Jun Mei Dongchen Qi Ting Liao Ziqi Sun School of Chemistry and Physics Queensland University of Technology Centre for Materials science Queensland University of Technology School of Materials science and Engineering Shaanxi University of Science and Technology School of Mechanical Medical and Process Engineering Queensland University of Technology

Metal exsolution engineering has been regarded as a promising strategy for activating intrinsically inert perovskite oxide catalysts toward efficient oxygen evolution *** metal exsolution processes on perovskites are often achieved by using the reducing hydrogen gas; however,this is not effective for the relatively stable phase,such as Ruddlesden-Popper perovskite oxides.

关键词： exsolution oxygen evolution perovskite reconstruction Ruddlesden–Popper

Vector field path following for a micro flapping-wing robot

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

science China(Information sciences) 2024年第6期67卷 442-444页

作者： Haifeng HUANG Yingte LIU Tao NIU Chang WANG Yao ZOU Wei HE School of Intelligence science and technology University of Science and Technology Beijing Institute of Artificial Intelligence University of Science and Technology Beijing Key Laboratory of Intelligent Bionic Unmanned Systems Ministry of EducationUniversity of Science and Technology Beijing

Flapping-wing robots mimic the flight patterns of natural flying creatures,granting them natural *** hold promising applications in the military reconnaissance,urban surveillance,and various other fields[1–3].However,the current limitations in size often restrict the suitability of existing bird-like flapping-wing robots,

关键词：

Neural network study of the nuclear ground-state spin distribution within a random interaction ensemble

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Nuclear science and Techniques 2024年第3期35卷 216-227页

作者： Deng Liu Alam Noor A Zhen-Zhen Qin Yang Lei School of Mathematics and Physics Southwest University of Science and TechnologyMianyang 621010China School of Nuclear science and technology Southwest University of Science and TechnologyMianyang 621010China

The distribution of the nuclear ground-state spin in a two-body random ensemble(TBRE)was studied using a general classification neural network(NN)model with two-body interaction matrix elements as input features and the corresponding ground-state spins as labels or output *** quantum many-body system problem exceeds the capability of our optimized NNs in terms of accurately predicting the ground-state spin of each sample within the ***,our NN model effectively captured the statistical properties of the ground-state spin because it learned the empirical regularity of the ground-state spin distribution in TBRE,as discovered by physicists.

关键词： Neural network Two-body random ensemble Spin distribution of nuclear ground state

维普期刊数据库

A FAS approach for stabilization of generalized chained forms: part 1. Discontinuous control laws

在线全文

学校读者我要写书评

暂无评论

science China(Information sciences) 2024年第2期67卷 118-143页

作者： Guang-Ren DUAN Center for Control Theory and Guidance technology Harbin Institute of Technology Center for Control science and technology Southern University of Science and Technology

In this paper, a type of general nonholonomic systems is proposed, which contains both the Brockett’s two example systems, and their extended n-dimensional chained forms, as special cases. For the stabilization of such systems, a stabilizing controller is proposed based on the fully actuated system(FAS)approach, which is discontinuous at the origin but time-invariant when the open-loop system is time-invariant,and drives the feasible trajectories of the system to the origin exponentially. Furthermore, the proposed FAS approach is also extended to the sub-normal system case and the time-delay system case.

关键词： nonholonomic systems feedback stabilization exponential convergence sub-fully actuated systems time-delay systems

HEN:a novel hybrid explainable neural network based framework for robust network intrusion detection

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

science China(Information sciences) 2024年第7期67卷 72-90页

作者： Wei WEI Sijin CHEN Cen CHEN Heshi WANG Jing LIU Zhongyao CHENG Xiaofeng ZOU School of Computer science and Engineering Xi'an University of TechnologyShaanxi Key Laboratory for Network Computing and Security Technology School of Computer science and technology Wuhan University of Science and Technology School of Future technology South China University of Technology Shenzhen Research Institute of Hunan university School of Computer science Hunan University of Technology and Business Institute for Infocomm Research (I2R) Agency for Science Technology and Reseach (A*STAR)

With the rapid development of network technology and the automation process for 5G, cyberattacks have become increasingly complex and threatening. In response to these threats, researchers have developed various network intrusion detection systems(NIDS) to monitor network traffic. However, the incessant emergence of new attack techniques and the lack of system interpretability pose challenges to improving the detection performance of NIDS. To address these issues, this paper proposes a hybrid explainable neural network-based framework that improves both the interpretability of our model and the performance in detecting new attacks through the innovative application of the explainable artificial intelligence(XAI)method. We effectively introduce the Shapley additive explanations(SHAP) method to explain a light gradient boosting machine(Light GBM) model. Additionally, we propose an autoencoder long-term short-term memory(AE-LSTM) network to reconstruct SHAP values previously generated. Furthermore, we define a threshold based on reconstruction errors observed during the training phase. Any network flow that surpasses the specified threshold is classified as an attack flow. This approach enhances the framework's ability to accurately identify attacks. We achieve an accuracy of 92.65%, a recall of 95.26%, a precision of 92.57%,and an F1-score of 93.90% on the dataset NSL-KDD. Experimental results demonstrate that our approach generates detection performance on par with state-of-the-art methods.

关键词： explainable artificial intelligence light gradient boosting machine machine learning network intrusion detection Shapley additive explanation hybrid explainable neural network (HEN)