检索结果-内蒙古大学图书馆

Probabilistic Graph Matching with Multiplicative Updating Algorithm for Correspondence between Remote Sensing Images

学校读者我要写书评

暂无评论

Probabilistic Graph Matching with Multiplicative Updating Al...

Information Communication and Software Engineering (ICICSE), IEEE International Conference on

作者： Jing Yang Xu Yang Zhang-Bing Zhou Zhi-Yong Liu School of Information Engineering China University of Geosciences(Beijing) State Key Laboratory of Management and Control for Complex Systems Institute of Automation Chinese Academy of Sciences Beijing China State Key Laboratory of Management and Control for Complex Systems Institute of Automation Chinese Academy of Sciences Beijing China School of Information Engineering China University of Geosciences(Beijing) TELECOM SudParis Beijing China State Key Laboratory of Management and Control for Complex Systems Institute of Automation Chinese Academy of Sciences Center for Excellence in Brain Science and Intelligence Technology Beijing China

ISBN: (纸本)9781665430302

Point correspondence is an important problem in remote sensing images matching task which lays the foundation for the image registration task. Due to the existence of the rotation transformation and the large scale, and the change of scenery between the two obtained images, the point correspondence problem is still challenging. Tradition algorithms such as the SIFT and SURF aim at using the point appearance to find the point correspondence. However, when the above situations happens, the appearance based algorithms will fail to solve the problem. In this paper, the structural information between points is introduced to the correspondence problem and a spectral graph matching algorithm is proposed based on the probabilistic scheme and the multiplicative updating scheme. The proposed algorithm has three advantages compared with the traditional algorithms. First, incorporating the structural cues to the matching problem makes it more robust to the large scale and the geometric transformation. Second, the best assignment between points which is interpreted by a probabilistic manner is figured out by ranking the probabilities. Third, the optimization problem is solved by an approximate spectral method efficiently and a multiplicative updating algorithm is introduced to ensure the final discrete solution. Experiments on the real remote sensing images validate the effectiveness of the proposed algorithm.

关键词： Image registration Software algorithms Approximation algorithms Probabilistic logic Sensors Internet Task analysis

EiHi Net: Out-of-Distribution Generalization Paradigm

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Wei, Qinglai Yuan, Beiming Chen, Diancheng State Key Laboratory for Management and Control of Complex Systems Institute of Automation Chinese Academy of Sciences Beijing100190 China School of Artificial Intelligence University of Chinese Academy of Sciences Beijing100049 China Institute of Systems Engineering Macau University of Science and Technology 999078 China

This paper develops EiHi to solve the out of distribution (O.o.D.) problem in recognition, especially the distribution shift of background. EiHi net is a model learning paradigm that can be blessed on any visual backbone. Different from the traditional deep model, namely find out correlated relationship between sample features and corresponding categories, EiHi net suffers less from spurious correlations. Fusing SimCLR and VIC-Reg via explicit and dynamical establishment of original - positive - negative sample pair as a minimal learning element, EiHi net iteratively establishes a relationship close to the causal one between features and labels, while suppresses the spurious correlations. To further reinforce the ability of the proposed model on causal reasoning, we develop a human-in-the-loop strategy, with few guidance samples, to prune the representation space directly. Finally, it is shown that the developed EiHi net makes significant improvements in the most difficult and typical O.o.D. dataset NICO, compared with the current SOTA results, without any domain (e.g. background) information. Copyright © 2022, The Authors. All rights reserved.

关键词： Machine learning

MULTI-VIEWPOINT AND MULTI-EVALUATION WITH FELICITOUS INDUCTIVE BIAS BOOST MACHINE ABSTRACT REASONING ABILITY

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Wei, Qinglai Chen, Diancheng Yuan, Beiming State Key Laboratory for Management and Control of Complex Systems Institute of Automation Chinese Academy of Sciences School of Artificial Intelligence University of Chinese Academy of Sciences Beijing China School of Artificial Intelligence University of Chinese Academy of Sciences Beijing China

Great endeavors have been made to study AI's ability in abstract reasoning, along with which different versions of RAVEN's progressive matrices (RPM) are proposed as benchmarks. Previous works give inkling that without sophisticated design or extra meta-data containing semantic information, neural networks may still be indecisive in making decisions regarding the RPM problems, after relentless training. Evidenced by thorough experiments, we show that, neural networks embodied with felicitous inductive bias, intentionally design or serendipitously match, can solve the RPM problems efficiently, without the augment of any extra meta-data. Our work also reveals that multiviewpoint with multi-evaluation is a key learning strategy for successful reasoning. Nevertheless, we also point out the unique role of meta-data by showing that a pre-training model supervised by the meta-data leads to a RPM solver with better performance. Source code can be found in https://***/QinglaiWeiCASIA/RavenSolver. Copyright © 2022, The Authors. All rights reserved.

关键词： Matrix algebra

Screening of metal-modified biochars for practical phosphorus recovery

学校读者我要写书评

暂无评论

Science of the Total Environment 2024年 956卷 177342页

作者： Li, Man Xie, Qian Xu, Fangxi Zhang, Yan Zhuang, Zechao Xu, Jiawei Xiang, Hai Li, Yongfu Cai, Yanjiang Chen, Zuliang Yu, Bing State Key Laboratory of Subtropical Silviculture Zhejiang A&F University Hangzhou311300 China Zhejiang A&F University Hangzhou311300 China Zhejiang Taizhou Ecological and Environmental Monitoring Center Taizhou318000 China Department of Chemistry Tsinghua University Beijing100084 China Jiangsu Key Laboratory of Numerical Simulation of Large Scale Complex Systems and School of Chemistry and Materials Science Nanjing Normal University Nanjing210023 China Fujian Key Laboratory of Pollution Control and Resource Reuse School of Environmental and Resource Sciences Fujian Normal University Fujian Province Fuzhou350007 China

The utilization of metal-modified biochars (MBCs) for practical phosphorus recovery has attracted significant research interest recently. However, the optimal choice of metals and modification methods for MBCs remains unclear. This study addresses this gap by comparing the phosphate adsorption capabilities of various MBCs using real municipal wastewater. The results show that zinc-modified biochar exhibits superior phosphate adsorption compared to biochars modified with calcium, magnesium, aluminum, and iron. Specifically, zinc-modified biochar prepared through metal-mediated biomass pyrolysis with alkaline soaking (ZnBC-OH) demonstrates the highest adsorption capacity, achieving 36.6 mg P/g in wastewater with a phosphate concentration of 5 mg P/L. This performance surpasses that of previously reported non-lanthanide modified biochars and is comparable to lanthanide-modified biochars. Mechanistic investigations reveal that the exceptional performance of ZnBC-OH is due to the presence of highly dispersed ZnO sites, which facilitate the formation of Zn3(PO4)2·4H2O precipitation, effectively retaining phosphate. Furthermore, a techno-economic analysis indicates that using ZnBC-OH in a fixed-bed column system can reduce phosphate levels from 6 mg L−1 to below 0.5 mg L−1 at a cost of 1.834 USD per ton of secondary treated wastewater, underscoring its promising application potential. © 2024 Elsevier B.V.

关键词： Wastewater treatment

Traffic Signal control Using Offline Reinforcement Learning

学校读者我要写书评

暂无评论

Traffic Signal Control Using Offline Reinforcement Learning

Chinese Automation Congress (CAC)

作者： Xingyuan Dai Chen Zhao Xiaoshuang Li Xiao Wang Fei-Yue Wang School of Artificial Intelligence University of Chinese Academy of Sciences The State Key Laboratory for Management and Control of Complex Systems Institute of Automation Chinese Academy of Sciences

ISBN: (纸本)9781665426480

The problem of traffic signal control is essential but remains unsolved. Some researchers use online reinforcement learning, including the off-policy one, to derive an optimal control policy through interaction between agents and environments in simulation. However, it is difficult to deploy the policy in real transportation systems due to the gap between simulated and real traffic data. In this paper, we consider an offline reinforcement learning method to tackle the problem. First, we construct a realistic traffic environment and obtain offline data based on a classic actuated traffic signal controller. Then, we use an offline reinforcement learning algorithm, namely conservative Q-learning, to learn an efficient control policy via offline datasets. We conduct experiments on a typical road intersection and compare the conservative Q-learning policy with the actuated policy and two data-driven policies based on off-policy reinforcement learning and imitation learning. Empirical results indicate that in the offline-learning setting the conservative Q-learning policy performs significantly better than other baselines, including the actuated policy, but the other two data-driven policies perform poorly in test scenarios.

关键词： Q-learning Automation Roads Transportation Optimal control Cloning Data models

Design of a curled hyper-redundant manipulator and the motion control with tip-following algorithm

学校读者我要写书评

暂无评论

Design of a curled hyper-redundant manipulator and the motio...

Chinese control and Decision Conference, CCDC

作者： Aoshun Zhang En Li Feng Zhang Rui Guo Mingrui Luo Yuwei Zhang School of Artificial Intelligence University of Chinese Academy of Sciences Beijing China The State Key Laboratory of Management and Control for Complex Systems Institute of Automation Chinese Academy of Sciences Beijing China Engineering Laboratory for Intelligent Industrial Vision CAS Beijing China State Grid Intelligent Technology Co. Ltd. Shandong China State Grid Shandong Electric Power Company Jinan Power Supply Company Shandong China

In order to meet the operational needs in some narrow environments, the hyper-redundant snake-shaped manipulator has received extensive attention and has been widely studied due to its compact shape and multiple degrees of freedom. At present, the research of hyper-redundant manipulator mainly focuses on the straight push platform. Considering the limitation of working space, this paper designed a hyper-redundant manipulator delivered by a curled turntable. According to the characteristics of turntable delivery, a tip-following algorithm based on curve fitting and a tip-following strategy based on terminal posture record are proposed. Finally, with the algorithm applied to the physical robot we made, we successfully realized the tip-following motion control of the hyper-redundant manipulator.

关键词：

Simulating unsteady fluid flows on a superconducting quantum processor

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Meng, Zhaoyuan Zhong, Jiarun Xu, Shibo Wang, Ke Chen, Jiachen Jin, Feitong Zhu, Xuhao Gao, Yu Wu, Yaozu Zhang, Chuanyu Wang, Ning Zou, Yiren Zhang, Aosai Cui, Zhengyi Shen, Fanhao Bao, Zehang Zhu, Zitian Tan, Ziqi Li, Tingting Zhang, Pengfei Xiong, Shiying Li, Hekang Guo, Qiujiang Wang, Zhen Song, Chao Wang, H. Yang, Yue State Key Laboratory for Turbulence and Complex Systems College of Engineering Peking University Beijing100871 China School of Physics ZJU-Hangzhou Global Scientific and Technological Innovation Center Zhejiang Key Laboratory of Micro-nano Quantum Chips and Quantum Control Zhejiang University Hangzhou310027 China Department of Engineering Mechanics School of Aeronautics and Astronautics Zhejiang University Hangzhou310027 China HEDPS-CAPT Peking University Beijing100871 China

Recent advancements of intermediate-scale quantum processors have triggered tremendous interest in the exploration of practical quantum advantage. The simulation of fluid dynamics, a highly challenging problem in classical physics but vital for practical applications, emerges as a good candidate for showing quantum utility. Here, we report an experiment on the digital simulation of unsteady flows, which consists of quantum encoding, evolution, and detection of flow states, with a superconducting quantum processor. The quantum algorithm is based on the Hamiltonian simulation using the hydrodynamic formulation of the Schrödinger equation. With the median fidelities of 99.97% and 99.67% for parallel single- and two-qubit gates respectively, we simulate the dynamics of a two-dimensional (2D) compressible diverging flow and a 2D decaying vortex with ten qubits. The experimental results well capture the temporal evolution of averaged density and momentum profiles, and qualitatively reproduce spatial flow fields with moderate noises. This work demonstrates the potential of quantum computing in simulating more complex flows, such as turbulence, for practical applications. Copyright © 2024, The Authors. All rights reserved.

关键词： Flow of fluids

Greenhouse Environment control based on Computational Experiments 20

学校读者我要写书评

暂无评论

Greenhouse Environment Control based on Computational Experi...

4th High Performance Computing and Cluster Technologies Conference, HPCCT 2020 and the 3rd International Conference on Big Data and Artificial Intelligence, BDAI 2020

作者： Weng, Yuchen Wang, Xiujuan Hua, Jing Wang, Haoyu Kang, Mengzhen State Key Laboratory of Management and Control for Complex Systems Institute of Automation Chinese Academy of Sciences School of Artificial Intelligence University of Chinese Academy of Sciences Haidian Beijing China

ISBN: (纸本)9781450375603

Greenhouse environment control is challenging since it is a partially open system being complex, nonlinear and highly uncertain. For such a complex system, empirical knowledge is often not sufficient to control the environment well. This paper proposes an approach for environment control based on computational experiments. The data source is from the second Autonomous Greenhouses International Challenge. Descriptive model is built with lightGBM algorithm for air temperature and CO2 concentration inside greenhouse. Computational experiments were carried out to optimize the ventilation opening, which has double effect of cooling and CO2 concentration decreasing. The aim of optimization is to keep the temperature as close as possible to the set line, while keeping a necessary level of CO2 concentration. The results indicate that the temperature can be better controlled than the original empirical methods for different conditions, while the CO2 concentration is maintained. The proposed approach overcomes the shortage of controlling ventilation with static knowledge, and can be applied to the real greenhouse. © 2020 ACM.

关键词： Greenhouses

Conditional Goal-oriented Trajectory Prediction for Interacting Vehicles with Vectorized Representation

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Li, Ding Zhang, Qichao Lu, Shuai Pan, Yifeng Zhao, Dongbin School of Artificial Intelligence University of Chinese Academy of Sciences Beijing100049 China State Key Laboratory of Management and Control for Complex Systems Institute of Automation Chinese Academy of Sciences Beijing100190 China Baidu Inc. Beijing100085 China

Predicting joint future trajectories of two interacting vehicles faces great challenges, due to the high degree of multimodality and uncertainty in the future interaction process. For the traditional joint prediction models, each vehicle is treated as an individual agent, and its multimodal future modes can be learned from an efficient Goal-oriented Trajectory Prediction (GTP) method. However, such marginal methods produce multimodal self-consistent trajectories over individual agents without considering their future interactions. Motivated by this, this paper aims to tackle the interactive behavior prediction task, and proposes a novel Conditional Goal-oriented Trajectory Prediction (CGTP) framework to jointly generate scene-compliant trajectories of two interacting agents. Our CGTP framework is an end-to-end and interpretable model, including three main stages: context encoding, goal interactive prediction and trajectory interactive prediction. First, a Goals-of-Interest Network (GoINet) is designed to extract the interactive features between agent-to-agent and agent-to-goals using a graph-based vectorized representation. Further, the Conditional Goal Prediction Network (CGPNet) focuses on goal interactive prediction via a combined form of marginal and conditional goal predictors. Finally, the Goal-oriented Trajectory Forecasting Network (GTFNet) is proposed to implement trajectory interactive prediction via the conditional goal-oriented predictors, with the predicted future states of the other interacting agent taken as inputs. In addition, a new goal interactive loss is developed to better learn the joint probability distribution over goal candidates between two interacting agents. In the end, the proposed method is conducted on Argoverse motion forecasting dataset, In-house cut-in dataset, and Waymo open motion dataset. The comparative results demonstrate the superior performance of our proposed CGTP model than the mainstream prediction methods. Copyright © 202

关键词： Forecasting