检索结果-内蒙古大学图书馆

Multimodal unified attention networks for vision-and-language interactions

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Yu, Zhou Cui, Yuhao Yu, Jun Tao, Dacheng Tian, Qi Key Laboratory of Complex Systems Modeling and Simulation School of Computer Science and Technology Hangzhou Dianzi University China UBTECH Sydney AI Centre School of Computer Science Faculty of Engineering University of Sydney Australia Noah's Ark Lab Huawei China

Learning an effective attention mechanism for multimodal data is important in many vision-and-language tasks that require a synergic understanding of both the visual and textual contents. Existing state-of-the-art approaches use co-attention models to associate each visual object (e.g., image region) with each textual object (e.g., query word). Despite the success of these co-attention models, they only model inter-modal interactions while neglecting intra-modal interactions. Here we propose a general 'unified attention' model that simultaneously captures the intra- and inter-modal interactions of multimodal features and outputs their corresponding attended representations. By stacking such unified attention blocks in depth, we obtain the deep Multimodal Unified Attention Network (MUAN), which can seamlessly be applied to the visual question answering (VQA) and visual grounding tasks. We evaluate our MUAN models on two VQA datasets and three visual grounding datasets, and the results show that MUAN achieves top level performance on both tasks without bells and whistles. Copyright © 2019, The Authors. All rights reserved.

关键词： Deep learning

Decentralized Event-Triggered Online Learning for Safe Consensus of Multi-Agent systems with Gaussian Process Regression

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Dai, Xiaobing Yang, Zewen Xu, Mengtian Liu, Fangzhou Hattab, Georges Hirche, Sandra TUM School of Computation Information and Technology Technical University of Munich Munich80333 Germany Robert Koch Institute Berlin Germany National Key Laboratory of Modeling and Simulation for Complex Systems Harbin Institute of Technology 150001 China Department of Mathematics and Computer Science Freie Universität Berlin Berlin Germany

Consensus control in multi-agent systems has received significant attention and practical implementation across various domains. However, managing consensus control under unknown dynamics remains a significant challenge for control design due to system uncertainties and environmental disturbances. This paper presents a novel learning-based distributed control law, augmented by an auxiliary dynamics. Gaussian processes are harnessed to compensate for the unknown components of the multi-agent system. For continuous enhancement in predictive performance of Gaussian process model, a data-efficient online learning strategy with a decentralized event-triggered mechanism is proposed. Furthermore, the control performance of the proposed approach is ensured via the Lyapunov theory, based on a probabilistic guarantee for prediction error bounds. To demonstrate the efficacy of the proposed learning-based controller, a comparative analysis is conducted, contrasting it with both conventional distributed control laws and offline learning methodologies. Copyright © 2024, The Authors. All rights reserved.

关键词： Multi agent systems

A PREA and OODA Computing Framework of Unmanned systems to Perform complex Tasks

学校读者我要写书评

暂无评论

IEEE Transactions on Intelligent Vehicles

IEEE Transactions on Intelligent Vehicles 2024年 1-18页

作者： Yang, Dongsheng Yu, Lei Zhang, Yi Zhu, Fenghua Li, Qiang Cui, Huachao Ma, Qiongmin Fan, Lili Tian, Yonglin Jinan University Guangzhou China Intelligent Science & Technology Academy Limited of CASIC Beijing China State Key Laboratory of Multimodal Artificial Intelligence Systems Institute of Automation Chinese Academy of Sciences Beijing China State Key Laboratory for Management and Control of Complex Systems Institute of Automation Chinese Academy of Sciences Beijing China 28th Research Institute of China Electronics Technology Group Corporation Nanjing China National Key Laboratory for Complex Systems Simulation National University of Defense Technology Changsha China School of Information and Electronics Beijing Institute of Technology Beijing China

A C2 computing framework for unmanned systems to perform complex tasks is proposed using methods of system of systems engineering, which is formed by combining the macro-scale command and control (C2) process mechanism model (PREA loop) and micro-scale C2 process mechanism model (OODA loop). Guided by PREA loop and OODA loop, the computing framework is divided into four steps and three kinds of transformations. The four steps are design, construction, operation monitoring, and assessment. The three transformations are tactical feedback, campaign feedback, and strategic feedback. A C2 organization model of joint landing combat force is established against the background of a typical complex task of joint landing operation, and the C2 computing framework based on the PREA &OODA is used to implement the C2 activities of the joint landing combat force. The influence of the computing framework on the performance of unmanned systems under different task conditions is verified, including task coordination load, task efficiency and sensitivity of unmanned systems response. Authors

关键词： Feedback

A Comprehensive Evaluation Framework for Deep Model Robustness

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Guo, Jun Bao, Wei Wang, Jiakai Ma, Yuqing Gao, Xinghai Xiao, Gang Liu, Aishan Dong, Jian Liu, Xianglong Wu, Wenjun State Key Lab of Software Development Environment Beihang University Beijing China China Electronics Standardization Institute Beijing China Zhongguancun Laboratory Beijing China Institute of Unmanned System Beihang University Beijing China National Key Laboratory for Complex Systems Simulation Beijing China

Deep neural networks (DNNs) have achieved remarkable performance across a wide range of applications, while they are vulnerable to adversarial examples, which motivates the evaluation and benchmark of model robustness. However, current evaluations usually use simple metrics to study the performance of defenses, which are far from understanding the limitation and weaknesses of these defense methods. Thus, most proposed defenses are quickly shown to be attacked successfully, which results in the "arm race" phenomenon between attack and defense. To mitigate this problem, we establish a model robustness evaluation framework containing 23 comprehensive and rigorous metrics, which consider two key perspectives of adversarial learning (i.e., data and model). Through neuron coverage and data imperceptibility, we use data-oriented metrics to measure the integrity of test examples;by delving into model structure and behavior, we exploit model-oriented metrics to further evaluate robustness in the adversarial setting. To fully demonstrate the effectiveness of our framework, we conduct large-scale experiments on multiple datasets including CIFAR-10, SVHN, and Im-ageNet using different models and defenses with our open-source platform. Overall, our paper provides a comprehensive evaluation framework, where re-searchers could conduct comprehensive and fast evaluations using the open-source toolkit, and the analytical results could inspire deeper understanding and further improvement to the model robustness. Copyright © 2021, The Authors. All rights reserved.

关键词： Deep neural networks

Theoretical analysis and simulation of phase separation in a driven bidirectional two-lane system

学校读者我要写书评

暂无评论

Physical Review E 2019年第3期100卷 032133-032133页

作者： Qing-Yi Hao Rui Jiang Mao-Bin Hu Yunxin Zhang Chao-Yun Wu Ning Guo Key Laboratory of Modeling Simulation and Control of Complex Ecosystem in Dabie Mountains of Anhui Higher Education Institutes School of Mathematics and Computational Science Anqing Normal University Anqing 246133 China School of Mathematical Sciences Fudan University Shanghai 200433 China MOE Key Laboratory for Urban Transportation Complex Systems Theory and Technology Beijing Jiaotong University Beijing 100044 China School of Engineering Science University of Science and Technology of China Hefei 230026 China School of Automotive and Transportation Engineering Hefei University of Technology Hefei 230009 China

The two-lane driven system is a type of important model to research some transport systems, and also a powerful tool to investigate properties of nonequilibrium state systems. This paper presents a driven bidirectional two-lane model. The dynamic characteristics of the model with periodic boundary are investigated by Monte Carlo simulation, simple mean field, and cluster mean field methods, respectively. By simulations, phase separations are observed in the system with some values of model parameters. When the phase separation does not occur, cluster mean field results are in good agreement with simulation results. According to the cluster mean field analysis and simulations, a conjecture about the condition that the phase separation happens is proposed. Based on the conjecture, the phase boundary distinguishing phase separation state and homogeneous state is determined, and a corresponding phase diagram is drawn. The conjecture is validated through observing directly the spatiotemporal diagram and investigating the coarsening process of the system by simulation, and a possible mechanism causing the phase separation is also discussed. These outcomes maybe contribute to understand deeply transport systems including the congestion and efficiency of the transport, and enrich explorations of nonequilibrium state systems.

关键词： Dynamical phase transitions Phase diagrams Phase separation Phase transitions Transport phenomena

Research on Image Stitching Methods for Cmos Grayscale Cameras in Industrial Scenarios

学校读者我要写书评

暂无评论

SSRN

SSRN 2024年

作者： Liu, Qi Huo, Ju Tang, Xiyu Xue, Muyao School of Electrical Engineering and Automation Harbin Institute of Technology Harbin150001 China National Key Laboratory of Modeling and Simulation for Complex Systems Heilongjiang Harbin150000 China School of Astronautics Harbin Institute of Technology Harbin150001 China Space Propulsion Technology Research Institute Shanghai Academy of Spaceflight Technology Shanghai201109 China

Addressing the limited field of view of CMOS grayscale cameras, complex lighting conditions, and the paucity of image features in industrial applications, this paper proposes a novel image stitching method for CMOS grayscale cameras under varying lighting conditions. It extends the camera's field of view while maintaining the interpretability of image features, thus enhancing the robustness and generalizability of image stitching across diverse lighting environments and feature-sparse settings. In the feature extraction phase, a hybrid deep feature extraction network is designed. By employing a deep learning-based approach for feature extraction, the network ensures an ample quantity of features. Building upon this, a line feature selection and reconstruction method is developed to further refine feature matching accuracy, increasing the number of matching lines under extreme lighting and feature-scarce situations, thereby enriching the image features for subsequent stitching steps. The subsequent image transformation phase introduces planar feature constraints;matching feature points and lines generate planar features, addressing alterations in the collective shape of planes common in industrial image stitching. The paper concludes with quantitative evaluation metrics for planar feature-based stitching. Experimental results validate the effectiveness and feasibility of our approach for image stitching of CMOS grayscale cameras under varied lighting conditions and in feature-deficient industrial settings. It offers a viable solution to the challenges of limited imaging fields of view faced by CMOS grayscale cameras in industrial applications. © 2024, The Authors. All rights reserved.

关键词： Feature extraction

A system-of-systems architecture-driven modeling method for combat system effectiveness simulation

学校读者我要写书评

暂无评论

A system-of-systems architecture-driven modeling method for ...

IEEE International Symposium on systems Engineering (ISSE)

作者： Xiaobo Li Weiping Wang Zhe Shu Ning Zhu Hua He Tianjun Liao Beijing Institute of System Engineering Beijing CN College of Information Systems and Management National University of Defense Technology Changsha Hunan P. R. China State Key Laboratory of Complex System Simulation Beijing Institute of Systems engineering Beijing China

ISBN: (纸本)9781509007943

Under the background of system-of-systems (SoS) counterwork, it is more apparent that the simulation modeling of the equipment should be domain-specific, formal, automatic and composable. Current equipment effectiveness simulation modeling approaches are mainly based on the methods and techniques from the modeling and simulation (M&S) community and software engineering community, and they are not capable to satisfy these four requirements systematically. Thus the systems engineering architecture model and domain knowledge should be used to support effectiveness simulation modeling more effectively. In this research, we propose an architecture-driven modeling method which conforms to the principle of architecture-driven development, uses ontology techniques to build equipment SoS architecture model and sub-domain ontology. This method performs architecture driven simulation modeling to realize the transformation from architecture models to simulation model frameworks; employs ontological metamodeling to design domain-specific modeling languages (DSML) based on the comprehensive usage of architecture models, sub-domain ontologies and M&S formalisms; integrates domain-specific simulation models from various domains using the model framework, and supports the composable development of simulation applications.

关键词： Computational modeling Ontologies Computer architecture Analytical models Mathematical model

Reputation-Based Conditional Investment Enhances the Evolution of Cooperation in Spatial Public Goods Game

学校读者我要写书评

暂无评论

Reputation-Based Conditional Investment Enhances the Evoluti...

International Conference on Identification, Information and Knowledge in the Internet of Things (IIKI)

作者： Hong Ding Lin Cao Hongjun Qiu Zhangbing Zhou Lei Shu Yizhi Ren Key Laboratory of Complex Systems Modeling and Simulation Ministry of Education China Hangzhou Dianzi University Hangzhou China School of Computer Science and Technologv Hangzhou Dianzi University Hangzhou China School of Information Engineering China University of Geosciences Beijing China Guangdong Petrochemical Equipment Fault Diagnosis Key Laboratory Guangdong University of Petrochemical Technology Maoming China

Most previous studies involving public goods games are investigated under a simplifying assumption that participators are compulsive in collective interactions and contribute unconditionally to the public pool. Nevertheless, how the conditional investment mechanism based on individual's reputation affects the evolution of cooperation in structured populations is still unclear. Here we introduce a reputation-based conditional investment rule for constituting participant groups into spatial threshold public goods game, where the public goods game can be conducted only if the participant number is not less than the threshold parameter. Interestingly, we find that large threshold parameter results in the optimal environment for cooperators' viability.

关键词： Games Investment Sociology Statistics Lattices Evolution (biology) Fault diagnosis

Singularity structure simplification of hexahedral mesh via weighted ranking

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Xua, Gang Ling, Ran Zhang, Yongjie Jessica Xiao, Zhoufang Ji, Zhongping Rabczuk, Timon School of Computer Science and Technology Hangzhou Dianzi University Hangzhou310018 China Key Laboratory of Complex Systems Modeling and Simulation Ministry of Education Hangzhou310018 China Department of Mechanical Engineering Carnegie Mellon University United States Institute of Structural Mechanics Bauhaus-Universitat Weimar Germany

In this paper, we propose an improved singularity structure simplification method for hexahedral (hex) meshes using a weighted ranking approach. In previous work, the selection of to-be-collapsed base complex sheets/chords is only based on their thickness, which will introduce a few closed-loops and cause an early termination of simplification and a slow convergence rate. In this paper, a new weighted ranking function is proposed by combining the valence prediction function of local singularity structure, shape quality metric of elements and the width of base complex sheets/chords together. Adaptive refinement and local optimization are also introduced to improve the uniformity and aspect ratio of mesh elements. Compared to thickness ranking methods, our weighted ranking approach can yield a simpler singularity structure with fewer base-complex components, while achieving comparable Hausdorff distance ratio and better mesh quality. Comparisons on a hex-mesh dataset are performed to demonstrate the effectiveness of the proposed method. Copyright © 2019, The Authors. All rights reserved.

关键词： Aspect ratio