检索结果-内蒙古大学图书馆

Direct Learning for Parameter-Varying Feedforward control: A Neural-Network Approach

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Kon, Johan van de Wijdeven, Jeroen Bruijnen, Dennis Tóth, Roland Heertjes, Marcel Oomen, Tom Control Systems Technology Group Departement of Mechanical Engineering Eindhoven University of Technology Eindhoven Netherlands ASML Veldhoven Netherlands Philips Engineering Solutions Eindhoven Netherlands Control Systems Group Electrical Engineering Eindhoven University of Technology Netherlands The Institute for Computer Science and Control Budapest Hungary Delft University of Technology Delft Netherlands

The performance of a feedforward controller is primarily determined by the extent to which it can capture the relevant dynamics of a system. The aim of this paper is to develop an input-output linear parameter-varying (LPV) feedforward parameterization and a corresponding data-driven estimation method in which the dependency of the coefficients on the scheduling signal are learned by a neural network. The use of a neural network enables the parameterization to compensate a wide class of constant relative degree LPV systems. Efficient optimization of the neural-network-based controller is achieved through a Levenberg-Marquardt approach with analytic gradients and a pseudolinear approach generalizing Sanathanan-Koerner to the LPV case. The performance of the developed feedforward learning method is validated in a simulation study of an LPV system showing excellent performance. © 2023, CC BY-NC-ND.

关键词： Learning systems

Excess Mean Squared Error of Empirical Bayes Estimators

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Ju, Yue Wahlberg, Bo Hjalmarsson, Håkan Division of Decision and Control Systems School of Electrical Engineering and Computer Science KTH Royal Institute of Technology Stockholm10044 Sweden Competence Centre for Advanced BioProduction by Continuous Processing AdBIOPRO Sweden

Empirical Bayes estimators are based on minimizing the average risk with the hyper-parameters in the weighting function being estimated from observed data. The performance of an empirical Bayes estimator is typically evaluated by its mean squared error (MSE). However, the explicit expression for its MSE is generally unavailable for finite sample sizes. To address this issue, we define a high-order analytical criterion: the excess MSE. It quantifies the performance difference between the maximum likelihood and empirical Bayes estimators. An explicit expression for the excess MSE of an empirical Bayes estimator employing a general data-dependent hyper-parameter estimator is derived. As specific instances, we provide excess MSE expressions for kernel-based regularized estimators using the scaled empirical Bayes, Stein’s unbiased risk estimation, and generalized cross-validation hyper-parameter estimators. Moreover, we propose a modification to the excess MSE expressions for regularized estimators for moderate sample sizes and show its improvement on accuracy in numerical simulations. Copyright © 2025, The Authors. All rights reserved.

关键词： Mean square error

The new notion of Bohl dichotomy for nonautonomous difference equations and its relation to exponential dichotomy

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Czornik, Adam Kitzing, Konrad Siegmund, Stefan Faculty of Automatic Control Electronics and Computer Science Silesian University of Technology Gliwice Poland Institute of Analysis Faculty of Mathematics TU Dresden Germany

In [7] the concept of Bohl dichotomy is introduced which is a notion of hyperbolicity for linear nonautonomous difference equations that is weaker than the classical concept of exponential dichotomy. In the class of systems with bounded invertible coefficient matrices which have bounded inverses, we study the relation between the set BD of systems with Bohl dichotomy and the set ED of systems with exponential dichotomy. It can be easily seen from the definition of Bohl dichotomy that ED ⊆ BD. Using a counterexample we show that the closure of ED is not contained in BD. The main result of this paper is the characterization int BD = ED. The proof uses upper triangular normal forms of systems which are dynamically equivalent and utilizes a diagonal argument to choose subsequences of perturbations each of which is constructed with the Millionshikov Rotation Method. An Appendix describes the Millionshikov Rotation Method in the context of nonautonomous difference equations as a universal *** Codes 37B55, 34D09 Copyright © 2023, The Authors. All rights reserved.

关键词： Difference equations

Direct Learning for Parameter-Varying Feedforward control: A Neural-Network Approach

学校读者我要写书评

暂无评论

Direct Learning for Parameter-Varying Feedforward Control: A...

IEEE Conference on Decision and control

作者： Johan Kon Jeroen van de Wijdeven Dennis Bruijnen Roland Tóth Marcel Heertjes Tom Oomen Departement of Mechanical Engineering Control Systems Technology Group Eindhoven University of Technology Eindhoven The Netherlands ASML Veldhoven The Netherlands Philips Engineering Solutions Eindhoven The Netherlands Control Systems Group Electrical Engineering Eindhoven University of Technology The Netherlands Institute for Computer Science and Control Budapest Hungary Delft University of Technology Delft The Netherlands

关键词：

UGNA-VPR: A Novel Training Paradigm for Visual Place Recognition Based on Uncertainty-Guided NeRF Augmentation

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Shen, Yehui Zhang, Lei Li, Qingqiu Zhao, Xiongwei Wang, Yue Lu, Huimin Chen, Xieyuanli College of Intelligence Science and Technology National University of Defense Technology China Faculty of Robot Science and Engineering Northeastern University China School of Electronic and Information Engineering Harbin Institute of Technology Shenzhen China State Key Laboratory of Industrial Control Technology Institute of Cyber-Systems and Control Zhejiang University China SIASUN Robot& Automation Co. Ltd China School of Computer Science Fudan University China

Visual place recognition (VPR) is crucial for robots to identify previously visited locations, playing an important role in autonomous navigation in both indoor and outdoor environments. However, most existing VPR datasets are limited to single-viewpoint scenarios, leading to reduced recognition accuracy, particularly in multi-directional driving or feature-sparse scenes. Moreover, obtaining additional data to mitigate these limitations is often expensive. This paper introduces a novel training paradigm to improve the performance of existing VPR networks by enhancing multi-view diversity within current datasets through uncertainty estimation and NeRF-based data augmentation. Specifically, we initially train NeRF using the existing VPR dataset. Then, our devised self-supervised uncertainty estimation network identifies places with high uncertainty. The poses of these uncertain places are input into NeRF to generate new synthetic observations for further training of VPR networks. Additionally, we propose an improved storage method for efficient organization of augmented and original training data. We conducted extensive experiments on three datasets and tested three different VPR backbone networks. The results demonstrate that our proposed training paradigm significantly improves VPR performance by fully utilizing existing data, outperforming other training approaches. We further validated the effectiveness of our approach on self-recorded indoor and outdoor datasets, consistently demonstrating superior results. Our dataset and code have been released at https://***/nubot-nudt/UGNA-VPR. Copyright © 2025, The Authors. All rights reserved.

关键词：

Boolean matrix compressed sensing

学校读者我要写书评

暂无评论

Boolean matrix compressed sensing

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Qiang Liu Mahdi Soleymani Hessam Mahdavifar National Laboratory of Industrial Control Technology Institute of Cyber-Systems and Control Zhejiang University Hangzhou China Halıcıoğlu Data Science Institute University of California San Diego La Jolla CA USA Department of Electrical and Computer Engineering Northeastern University Boston MA USA

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

In real-world datasets, leveraging the low-rank and sparsity properties enables developing efficient algorithms across a diverse array of data-related tasks, including compression, compressed sensing, matrix completion, etc. Notably, these two properties often coexist in certain real-world datasets, especially in Boolean datasets and quantized real-valued datasets. To harness the advantages of low-rank and sparsity simultaneously, we adopt a technique inspired by compressed sensing and Boolean matrix completion. Our approach entails compressing a low-rank sparse Boolean matrix by performing inner product operations with a randomly generated Boolean matrix. We then propose a decoding algorithms based on message-passing techniques to recover the original matrix. Our experiments demonstrate superior recovery performance of our proposed algorithms compared to Boolean matrix completion, with equal measurement requirements.

关键词： Image coding Signal processing algorithms Signal processing Acoustics Sensors Decoding Sparse matrices Speech processing Compressed sensing

OARecon: Object-Aware Viewpoint Augmentation for Indoor Compositional Reconstruction

学校读者我要写书评

暂无评论

OARecon: Object-Aware Viewpoint Augmentation for Indoor Comp...

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Yuanyuan Ding Yiming Fei Jiandang Yang Xiaobin Wei Jiajun Lv Yong Liu Institute of Cyber-Systems and Control Zhejiang University Hangzhou China College of Computer Science and Technology Zhejiang University Hangzhou China WASU Media & Network Co. Ltd Hangzhou China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Real-world scenes likely involve repetitive objects indicating that the reconstruction of the target object can be supplemented by the views of other identical objects. However, traditional 3D reconstruction methods do not take this a priori knowledge into account and fail to make full use of the available information. In this paper, we propose an object-aware viewpoint augmentation scheme for indoor compositional reconstruction. Within this scheme, a viewpoint supplementation strategy based on signed distance function and neural radiance fields is proposed to fully leverage the information from repetitive objects such that the occlusion problem is suppressed. Moreover, this scheme introduces monocular uncertainty priors and regional smoothness constraints to enhance the reconstruction accuracy of slender and thin structures and the smoothness of occluded background, respectively. Experimental results considering both synthetic and real-world scenes demonstrate that our method effectively improves the reconstruction quality of repetitive objects and background.

关键词： Uncertainty Accuracy Three-dimensional displays Pose estimation Buildings Signal processing Benchmark testing Neural radiance field Acoustics Speech processing

Hyperbolic Binary Neural Network

学校读者我要写书评

暂无评论

IEEE Transactions on Neural Networks and Learning Systems 2024年第6期36卷 10325-10333页

作者： Jun Chen Jingyang Xiang Tianxin Huang Xiangrui Zhao Yong Liu National Special Education Resource Center for Children with Autism Zhejiang Normal University Hangzhou China Institute of Cyber-Systems and Control Zhejiang University Hangzhou China School of Computer Science and Technology Zhejiang Normal University Jinhua China

Binary neural network (BNN) converts full-precision weights and activations into their extreme 1-bit counterparts, making it particularly suitable for deployment on lightweight mobile devices. While BNNs are typically formulated as a constrained optimization problem and optimized in the binarized space, general neural networks are formulated as an unconstrained optimization problem and optimized in the continuous space. This article introduces the hyperbolic BNN (HBNN) by leveraging the framework of hyperbolic geometry to optimize the constrained problem. Specifically, we transform the constrained problem in hyperbolic space into an unconstrained one in Euclidean space using the Riemannian exponential map. On the other hand, we also propose the exponential parametrization cluster (EPC) method, which, compared with the Riemannian exponential map, shrinks the segment domain based on a diffeomorphism. This approach increases the probability of weight flips, thereby maximizing the information gain in BNNs. Experimental results on CIFAR10, CIFAR100, and ImageNet classification datasets with VGGsmall, ResNet18, and ResNet34 models illustrate the superior performance of our HBNN over state-of-the-art methods.

关键词： Neural networks Vectors Optimization Geometry Manifolds Transforms Quantization (signal) Aerospace electronics Training Mobile handsets

Real Time Sound Source Localization Using von-Mises ResNet

学校读者我要写书评

暂无评论

Real Time Sound Source Localization Using von-Mises ResNet

2024 IEEE/SICE International Symposium on System Integration, SII 2024

作者： Bozkurtlar, Mert Yen, Benjamin Itoyama, Katsutoshi Nakadai, Kazuhiro School of Engineering Tokyo Institute of Technology Department of Systems and Control Engineering 2-12-1 Ookayama Meguro-ku Tokyo152-8552 Japan Istanbul Technical University Faculty of Management Department of Industrial Engineering Maçka/Beşiktaş/Harbiye Istanbul34367 Turkey Istanbul Technical University Faculty of Computer and Informatics Engineering Department of Computer Engineering Sariyer/Ayazaga Istanbul34485 Turkey The Nakadai Lab Japan Honda Research Institute Japan Co. Ltd 8-1 Honcho Saitama Wako351-0188 Japan

ISBN: (纸本)9798350312072

This paper addresses the task of learning periodic information using deep neural networks to achieve real-time, environment-independent sound source localization. Previous papers showed phase data is the most significant cue in sound source localization tasks and the proposed vM-B DNN was validated to be able to handle such periodic information using on synthesized data. However, they haven't shown its effectiveness and robustness in realistic use cases. This paper introduces a more complex model based on residual networks and adapts vM-B activation function for convolutional layers for use cases that require real-time predictions in dynamically changing environments. © 2024 IEEE.

关键词： Deep neural networks