Article
Computer Science, Information Systems
Xudong Zhang, Lei Wang
Summary: This paper focuses on robust sparse M-estimation over decentralized networks in the presence of Byzantine attacks. It proposes algorithms that are provably robust against Byzantine attacks and achieve linear convergence rates by combining robust aggregation rules, gradient tracking, and proximal algorithm.
INFORMATION SCIENCES
(2024)
Article
Automation & Control Systems
Chen Liu, Gang Wang, Xin Guan, Chutong Huang
Summary: In this paper, a framework that combines M-estimation and information-theoretic learning (ITL)-based Kalman filter under impulsive noises is proposed. The proposed framework inhibits the divergence trend of ITL-based Kalman filters at low kernel bandwidth and improves the performance at large kernel bandwidth. Monte Carlo simulations demonstrate the robustness and effectiveness of the proposed algorithm.
Article
Computer Science, Artificial Intelligence
Jiyuan Tu, Weidong Liu, Xiaojun Mao
Summary: This paper presents a Byzantine-resilient method for distributed sparse M-estimation. By constructing a pseudo-response variable and transforming the optimization problem, a communication-efficient distributed algorithm is developed. Theoretically, it is proven that the proposed method achieves fast convergence and a support recovery result is established.
Article
Geography, Physical
Yuting Zhao, Jungho Im, Zhen Zhen, Yinghui Zhao
Summary: Accurate quantification of individual tree parameters is crucial for precise forest inventory and sustainable forest management. The combination of terrestrial laser scanning (TLS) and unmanned aerial vehicle laser scanning (ULS) is an effective approach to overcome the limitations of capturing complete tree structure in dense forests. This study proposed optimized algorithms for registering TLS and ULS point data, which achieved high accuracy in individual tree crown delineation, DBH, and tree height estimations in different forest types.
GISCIENCE & REMOTE SENSING
(2023)
Article
Computer Science, Information Systems
Xinxin Zhang, Ronggang Wang, Da Chen, Yang Zhao, Wen Gao
Summary: This paper proposes a novel blind motion deblurring method for blurred images including light streaks, reducing the influence of outliers on deconvolution and utilizing more information to estimate the blur kernel. Experimental results demonstrate the high accuracy of the algorithm in restoring both synthetic and real images.
IEEE TRANSACTIONS ON MULTIMEDIA
(2021)
Article
Engineering, Electrical & Electronic
Wang Guangcai, Xiaosu Xu, Tao Zhang
Summary: The MMCKF is a method that combines M-M estimation and nonlinear CKF to enhance the performance of INS/GPS integration system, capable of addressing both the contamination of predicted state and measurement by non-Gaussian noise, as well as the issue of large initial attitude error.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Engineering, Electrical & Electronic
Zhongjin Xue, Shuo Cheng, Liang Li, Zhihua Zhong, Hongyuan Mu
Summary: This paper proposes a robust filter algorithm for vehicle state estimation with unknown driver steering torque. The algorithm suppresses outliers by transforming and linearizing the nonlinear model, and addresses uncertainties using an iterative algorithm. Comparative simulations and experiments with other commonly used filter algorithms demonstrate the effectiveness and robustness of the proposed algorithm.
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY
(2022)
Article
Geography, Physical
Jiayuan Li, Qingwu Hu, Yongjun Zhang, Mingyao Ai
Summary: This paper proposes a robust symmetric ICP (RSICP) algorithm to address the limitations of traditional ICP, including small convergence basin and sensitivity to outliers and partial overlaps. The algorithm introduces a new symmetric point-to-plane distance metric and an adaptive robust loss, as well as a simple and effective linearization method. Extensive experiments demonstrate that the proposed algorithm outperforms other methods in terms of both accuracy and efficiency.
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2022)
Article
Engineering, Electrical & Electronic
Christian A. Schroth, Michael Muma
Summary: This article explores robust Bayesian cluster enumeration techniques for handling heavy-tailed noise and outliers in data sets using Real Elliptically Symmetric (RES) distributed mixture models and M-estimators. The derived robust criterion is applied to datasets with finite sample size and provides an asymptotic approximation to reduce computational cost at large sample sizes. The algorithms demonstrate significant improvements in robustness compared to existing methods when applied to simulated and real-world data sets.
IEEE TRANSACTIONS ON SIGNAL PROCESSING
(2021)
Article
Multidisciplinary Sciences
Khalid Ul Islam Rather, Eda Gizem Kocyigit, Ronald Onyango, Cem Kadila
Summary: This article proposes a new robust ratio type estimator using the Uk's redescending M-estimator for estimating the finite population mean in the presence of outliers in the dataset. The mean square error equation of the proposed estimator is obtained and compared with traditional ratio-type estimators, robust regression estimators, and other existing redescending M-estimators. A real-life data and simulation study are conducted to demonstrate the efficiency of the proposed estimator. The results show that the proposed estimator outperforms other estimators in both simulation and real data studies.
Article
Environmental Sciences
Kejia Huang, Chenliang Wang, Wenjiao Shi
Summary: This paper proposes a rotation-invariant estimation method for high-precision geo-registration in AR maps. It improves the accuracy of generating heading data and eliminates alignment errors between geospatial data and the ground surface. The experimental results show that it achieves AR geo-registration accuracy at the 0.1-degree level, about twice as high as traditional methods.
Article
Computer Science, Artificial Intelligence
Hao Xiong, Junchi Yan, Zengfeng Huang
Summary: Skip-gram models are cost-effective in large-scale network embedding and closely related to word2vec and NCE. The differences among embedding methods lie in how the node neighborhood is modeled, while NCE methods commonly involve two basic NCE components. However, NCE-based objectives suffer from slow convergence speed and difficulty in capturing non-linearity of complex networks. To address these issues, we propose a distance-based term to be added to the NCE term, which effectively improves the performance in node classification and network reconstruction tasks.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)
Article
Automation & Control Systems
Qingwen Ma, Xin Xu, Ronghua Zhang, Quan Xiong, Xinglong Zhang, Xianjian Zhang
Summary: This article proposes a robust consensus control scheme based on the convergence rate estimation, aiming to accelerate the convergence rate and improve the control performance of multi-agent systems (MASs). A novel convergence rate indicator is defined to measure the influence of model nonlinearity and state couplings within and among agents on convergence rate. A distributed consensus controller with the indicator-based additive term is designed to enhance the robustness of the system. Theoretical analysis proves the system's robustness and stability. Simulation results validate the effectiveness of the proposed controller.
INTERNATIONAL JOURNAL OF ROBUST AND NONLINEAR CONTROL
(2023)
Article
Engineering, Electrical & Electronic
Tan-Jan Ho
Summary: This paper proposes new algorithms for robust mobile location estimation using range measurements from multiple base stations under probabilistically mixed line-of-sight/non-line-of-sight conditions. The algorithms address the issue of unknown NLOS statistics and occurrence probabilities, and achieve stable and satisfactory location estimates. Simulation results demonstrate their superiority over other methods in terms of performance.
IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Linlin Zha, Kai Ma, Guoqiang Li, Qi Fang, Xiaobin Hu
Summary: This study proposes a new approach called robust double-parallel extreme learning machine (RDELM), based on an improved M-estimation optimized double-parallel extreme learning machine, to solve the problem of improving the regression accuracy and model stability of the extreme learning machine (ELM). The proposed method constructs RD-ELM with a double parallel forward structure and uses an improved M-estimation to calculate the output weights of the neural network. Experimental results demonstrate that RD-ELM has strong anti-jamming ability in handling outliers and noise, and outperforms other methods in terms of robustness and generalization performance in various benchmark data and practical experiments.
ADVANCED ENGINEERING INFORMATICS
(2022)