Article
Computer Science, Interdisciplinary Applications
Ugur G. Abdulla, Roby Poteau
Summary: A numerical method for identifying parameters in large-scale systems of nonlinear ODEs in systems biology is introduced, combining optimization, sensitivity analysis, and regularization. The method demonstrates superlinear convergence in testing on canonical benchmark models, making it suitable for partial and noisy measurements. The developed software package qlopt shows advantages over popular methods/software like lsqnonlin, finincon, and nl2sol.
JOURNAL OF COMPUTATIONAL PHYSICS
(2021)
Article
Computer Science, Artificial Intelligence
Ning Ding, Yujia Qin, Guang Yang, Fuchao Wei, Zonghan Yang, Yusheng Su, Shengding Hu, Yulin Chen, Chi-Min Chan, Weize Chen, Jing Yi, Weilin Zhao, Xiaozhi Wang, Zhiyuan Liu, Hai-Tao Zheng, Jianfei Chen, Yang Liu, Jie Tang, Juanzi Li, Maosong Sun
Summary: With the rise of pre-trained language models (PLMs) and the pre-training-fine-tuning approach, it has become clear that larger models generally achieve better performance. However, scaling up PLMs leads to high costs and impracticality in terms of fine-tuning and storing all parameters. To address this, parameter-efficient adaptation of PLMs, known as delta-tuning, focuses on optimizing a subset of parameters while keeping the rest fixed, reducing computation and storage costs. This article discusses and analyzes the different approaches of delta-tuning and explores their correlations and differences, providing a unified categorization criterion. Theoretical principles underlying the effectiveness of delta-tuning are also discussed, along with an empirical study on numerous natural language processing tasks.
NATURE MACHINE INTELLIGENCE
(2023)
Article
Automation & Control Systems
Wenhua Shao, Haiyong Luo, Fang Zhao, Hui Tian, Jingyu Huang, Antonino Crivello
Summary: Traditional Wi-Fi-based floor identification methods often have reduced accuracy when applied in real large and multistorey environments. This article proposes an adaptive Wi-Fi-based floor identification algorithm that improves accuracy by leveraging Wi-Fi features and spatial similarity.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2022)
Article
Engineering, Civil
Yongbiao Chen, Sheng Zhang, Fangxin Liu, Chenggang Wu, Kaicheng Guo, Zhengwei Qi
Summary: This paper develops an efficient vehicle re-identification system (DVHN) that reduces memory usage and improves retrieval efficiency while maintaining retrieval accuracy for large-scale retrieval tasks. It learns discrete compact binary hashing codes for each image through deep hashing learning and optimizes the feature learning network and hash code generating module. Extensive experiments on two widely-studied vehicle re-identification datasets demonstrate the superiority of the method.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Automation & Control Systems
Jiarui Zhang, Steven X. Ding, Deyu Zhang, Linlin Li
Summary: This paper develops a distributed fault detection approach for large-scale interconnected systems using sensor networks. The one-step prediction based on measured data is implemented in a distributed fashion, allowing each node to receive corresponding estimations and innovation sequences in real-time. The innovation sequences are then used to improve the estimation result through filtering and smoothing, and to detect faults. A case study demonstrates the efficient performance of the distributed approach in fault detection.
IET CONTROL THEORY AND APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Amine Elhafsi, Rohan Sinha, Christopher Agia, Edward Schmerling, Issa A. D. Nesnas, Marco Pavone
Summary: As robots become more capable and encounter a wider range of environments, the risk of edge case failures or anomalies becomes ever present. This study explores the use of large language models to detect semantic anomalies in complex scenarios, demonstrating its effectiveness in identifying such cases and aligning with human reasoning.
Article
Management
Srikanth Jagabathula, Paat Rusmevichientong, Ashwin Venkataraman, Xinyi Zhao
Summary: We propose an efficient estimation method for large-scale tree logit models, utilizing a novel change-of-variables transformation to express the negative log-likelihood as a strictly convex function. Our algorithm computes a sequence of parameter estimates using simple closed-form updates, relying only on first-order information. Numerical results demonstrate that our approach outperforms state-of-the-art optimization methods, especially for large-scale tree logit models with thousands of nodes.
OPERATIONS RESEARCH
(2023)
Article
Automation & Control Systems
Vahab Rostampour, Riccardo M. G. Ferrari, Andre M. H. Teixeira, Tamas Keviczky
Summary: This article addresses two limitations in current distributed model based approaches for anomaly detection in large-scale uncertain nonlinear systems: the high conservativeness of detection thresholds and the requirement for different parties to regularly communicate local measurements. To address these issues, a novel set-based threshold and privacy-preserving mechanism are proposed to ensure robustness and privacy.
IEEE TRANSACTIONS ON AUTOMATIC CONTROL
(2021)
Article
Computer Science, Artificial Intelligence
Guojiang Xiong, Xuan Xie, Zixia Yuan, Xiaofan Fu
Summary: Fault section diagnosis is critical for the safe operation of power systems. This study proposes a method based on structure adaptive hierarchical extreme learning machines for diagnosing large-scale power systems. The method optimizes the HELM model using differential evolution algorithm and achieves higher accuracy and fault tolerance in identifying complex fault cases.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Review
Green & Sustainable Science & Technology
Jose Joaquin Aguilera, Wiebke Meesenburg, Torben Ommen, Wiebke Brix Markussen, Jonas Lundsted Poulsen, Benjamin Zuehlsdorf, Brian Elmegaard
Summary: This study provides a description of faults in the operation of large-scale heat pumps, categorizing them based on potential causes, mitigation methods, and detection and diagnosis approaches. Faults in the compressor, evaporator, and source heat exchanger were more frequently observed, with fouling of heat exchangers and refrigerant leakage being the most common. Future research directions include developing specific fault detection and diagnosis methods for large-scale heat pump applications and integrating performance degradation monitoring with fault detection and diagnosis.
RENEWABLE & SUSTAINABLE ENERGY REVIEWS
(2022)
Article
Computer Science, Artificial Intelligence
Deng-Ping Fan, Zheng Lin, Zhao Zhang, Menglong Zhu, Ming-Ming Cheng
Summary: This article makes contributions to RGB-D SOD by collecting a new SIP dataset, conducting a large-scale benchmark comparing contemporary methods, and proposing the D(3)Net model. D(3)Net outperforms prior contenders and can efficiently extract salient object masks for real scenes.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2021)
Article
Chemistry, Physical
Sara Sattarzadeh, Tanushree Roy, Satadru Dey
Summary: Thermal fault diagnostics in large format pouch or prismatic cells pose challenges due to the two-dimensional nature of temperature distribution. A framework is proposed to optimize sensor locations and design filtering schemes for improved detection and isolation of thermal faults. This framework is validated through experimental and simulation studies on a commercial battery cell.
JOURNAL OF POWER SOURCES
(2021)
Article
Mathematics, Applied
Zhen-Lei Ma, Xiao-Jian Li
Summary: This paper addresses the fault detection problem in large-scale network systems with unknown system dynamic matrices. By estimating the unmeasurable interconnection terms within a data-driven framework, designing a residual generator, and using an H-/H-infinity mixed optimization scheme, the proposed approach enhances the sensitivity to actuator faults and robustness against measurement noises. The advantages and effectiveness of the proposed method are verified through a numerical example.
APPLIED MATHEMATICS AND COMPUTATION
(2022)
Article
Automation & Control Systems
Mounira Hamdi, Lhassane Idomhgar, Samira Kamoun, Mondher Chaoui, Abdenaceur Kachouri
Summary: This paper proposes a recursive distributed parameter estimation algorithm based on the minimization of the prediction estimation error method for large-scale systems. The algorithm considers a class of large-scale systems composed of several interconnected sub-systems, each modeled by a linear discrete-time state space mathematical model with unknown parameters. Convergence analysis is achieved using the Lyapunov approach. The theoretical analysis and simulation results prove the effectiveness of the proposed algorithm.
IET CONTROL THEORY AND APPLICATIONS
(2023)
Article
Engineering, Electrical & Electronic
Martin Nachtsheim, Johannes Ernst, Christian Endisch, Ralph Kennel
Summary: The recursive least squares (RLS) algorithm has the potential to be used for innovative electric vehicle diagnosis based on online parameter identification (OPI) of induction machines (IMs). However, the algorithm faces challenges in the automotive environment due to special machine designs and highly dynamic operation in a wide speed and load range. This work compares different algorithm extensions and a novel RLS algorithm with multiple variable forgetting factors for IMs, analyzing their handling of transient parameter behavior. The real-time performance of the identification algorithms is given special attention for implementation in automotive embedded systems.
IEEE TRANSACTIONS ON TRANSPORTATION ELECTRIFICATION
(2023)
