检索结果-内蒙古大学图书馆

Condensed composite memory continual learning

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Wiewel, Felix Yang, Bin Institute of Signal Processing and System Theory University of Stuttgart Stuttgart Germany

Deep Neural Networks (DNNs) suffer from a rapid decrease in performance when trained on a sequence of tasks where only data of the most recent task is available. This phenomenon, known as catastrophic forgetting, prevents DNNs from accumulating knowledge over time. Overcoming catastrophic forgetting and enabling continual learning is of great interest since it would enable the application of DNNs in settings where unrestricted access to all the training data at any time is not always possible, e.g. due to storage limitations or legal issues. While many recently proposed methods for continual learning use some training examples for rehearsal, their performance strongly depends on the number of stored examples. In order to improve performance of rehearsal for continual learning, especially for a small number of stored examples, we propose a novel way of learning a small set of synthetic examples which capture the essence of a complete dataset. Instead of directly learning these synthetic examples, we learn a weighted combination of shared components for each example that enables a significant increase in memory efficiency. We demonstrate the performance of our method on commonly used datasets and compare it to recently proposed related methods and baselines. © 2021, CC BY.

关键词： Deep neural networks

Plasma induced dynamic coupling of microscopic factors to collaboratively promote EM losses coupling of transition metal dichalcogenide absorbers

学校读者我要写书评

暂无评论

Advanced Powder Materials

Advanced Powder Materials 2024年第3期3卷 1-11页

作者： Jiaming Wen Geng Chen Shengchong Hui Zijing Li Jijun Yun Xiaomeng Fan Limin Zhang Qian He Xingmin Liu Hongjing Wu MOE Key Laboratory of Material Physics and Chemistry under Extraordinary School of Physical Science and TechnologyNorthwestern Polytechnical UniversityXi'an710072China School of Material Science and Engineering Northwestern Polytechnical UniversityXi'an 710072China Frontiers Science Center for Deep Ocean Multispheres and Earth System and Key Laboratory of Marine Chemistry Theory and TechnologyMinistry of EducationOcean University of ChinaQingdao 266100China Institute of Materials Science Technische Universitӓt DarmstadtAlarich-Weiss-Str.264287DarmstadtGermany

Plasma as the fourth state of matter has attracted great attention for material surface modification,which could induce changes in material microscopic factors,such as defects,phase transitions,crystallinity,and so ***,the interactions among those microscopic factors and regulation mechanism of macroscopic properties have rarely been ***-dimensional(2D)transition metal dichalcogenide with tunable structure and phase is one of the most promising electromagnetic wave(EMW)absorbers,which provides a favorable platform for systematically studying the dynamic coupling of its microscopic ***,we constructed a NaBH_(4) solution-assisted Ar plasma method to modify the 2H-MoS_(2)and 1T-WS_(2)for exploring the regulation mechanism of microscopic *** MoS_(2)and WS_(2),NaBH_(4) solution-assisted Ar plasma treatment behaves with different effects on dielectric responses,realizing dynamic coupling of material microscopic factors to collaboratively promote EM losses ***,the MS-D3-0.5(MoS_(2),3 kV voltage,0.5 mol L^(-1)NaBH_(4) solution)displays an optimum effective absorption bandwidth of 8.01 GHz,which is 319.4%more than that of MS-raw *** study not only reveals the novel mechanism of plasma induced dynamic coupling of microscopic factors for EMW dissipation,but also presents a new method of plasma-dominated surface modification to optimize the EMW absorption performance.

关键词： TMDs Ar plasma Defect Metal single atom Dynamic coupling

Metal matrix nanocomposites in tribology: Manufacturing, performance, and mechanisms

学校读者我要写书评

暂无评论

Friction 2022年第10期10卷 1596-1634页

作者： Shuaihang PAN Kaiyuan JIN Tianlu WANG Zhinan ZHANG Long ZHENG Noritsugu UMEHARA Department of Mechanical and Aerospace Engineering University of California Los Angeles(UCLA)Los AngelesCA 90095USA Physical Intelligence Department Max Planck Institute for Intelligent SystemsStuttgart 70569Germany Stake Key Laboratory of Mechanical System and Vibration Shanghai Jiao Tong UniversityShanghai 200240China Key Laboratory of Bionic Engineering(Ministry of Education) College of Biological and Agricultural EngineeringJilin UniversityChangchun 130022China Micro-Nano Mechanical Science Laboratory Department of Micro-Nano Mechanical Science and EngineeringGraduate School of EngineeringNagoya UniversityChikisa-ku Furo-choNagoyaAichi 464-8601Japan

Metal matrix nanocomposites(MMNCs)become irreplaceable in tribology industries,due to their supreme mechanical properties and satisfactory tribological ***,due to the dual complexity of MMNC systems and tribological process,the anti-friction and anti-wear mechanisms are unclear,and the subsequent tribological performance prediction and design of MMNCs are not easily possible:A critical up-to-date review is needed for MMNCs in *** review systematically summarized the fabrication,manufacturing,and processing techniques for high-quality MMNC bulk and surface coating materials in ***,important factors determining the tribological performance(mainly anti-friction evaluation by the coefficient of friction(CoF)and anti-wear assessment with wear rate)in MMNCs have been investigated thoroughly,and the correlations have been analyzed to reveal their potential coupling/synergetic roles of tuning tribological behavior of *** importantly,this review combined the classical metal/alloy friction and wear theories and adapted them to give a(semi-)quantitative description of the detailed mechanisms of improved anti-friction and anti-wear performance in *** guarantee the universal applications of these mechanisms,their links with the analyzed influencing factors(e.g.,loading forces)and characteristic features like tribo-film have been *** approach forms a solid basis for understanding,predicting,and engineering MMNCs’tribological behavior,instead of pure phenomenology and experimental ***,the pathway to achieve a broader application for MMNCs in tribo-related fields like smart materials,biomedical devices,energy storage,and electronics has been concisely discussed,with the focus on the potential development of modeling,experimental,and theoretical techniques in MMNCs’tribological *** general,this review tries to elucidate the complex tribo-performances of MMNCs in a fundamentally universal yet straightfor

关键词： metal matrix nanocomposites nanophases tribology manufacturing processes anti-wear performance strengthening effects anti-wear mechanisms

A Concept for Myocardial Current Density Estimation with Magnetoelectric Sensors

学校读者我要写书评

暂无评论

Current Directions in Biomedical engineering 2023年第1期9卷 89-92页

作者： Schmidt, Gerhard Engelhardt, Erik Elzenheimer, Eric Hoffman, Johannes Schmidt, Tobias Zaman, Adrian Frey, Norbert Digital Signal Processing and System Theory Institute of Electrical Engineering and Information Technology Faculty of Engineering Kiel University Kaiserstr. 2 Kiel 24143 Germany Medical Department III Specialised in Cardiology and Angiology University Medical Center Schleswig-Holstein Arnold-Heller-Str. 3 / House K3 Kiel 24105 Germany Internal Medicine III: Heart Vascular and Lung University Medical Center Heidelberg Im Neuenheimer Feld 410 Heidelberg 69120 Germany

In this paper we present a novel noninvasive approach to estimate current densities in the heart from magnetocardiography. The proposed algorithm uses nested optimization to model current densities in equally-sized voxels of myocardial tissue. First-order Thiran all-pass filters are used to describe the propagation between *** demonstrate feasibility of the algorithm for a noise-free single-layer simulation. However, challenges remain, such as addressing measurement noise and optimizing propagation velocity. Overall, this approach has the potential to complement or replace invasive catheter-based electrophysiological studies for localization of arrhythmogenic tissue. © 2023 the author(s), published by Walter de Gruyter Berlin/Boston.

关键词： Magnetocardiography

Design of a Flexible Capture Mechanism Inspired by Sea Anemone for Non-cooperative Targets

学校读者我要写书评

暂无评论

Chinese Journal of Mechanical engineering 2021年第4期34卷 236-248页

作者： Jiankun Yang Chengwei Ren Chenghao Yang Youyu Wang Shumin Wan Rongjie Kang Key Laboratory of Mechanism Theory and Equipment Design of Ministry of Education Tianjin UniversityTianjin 300350China State Key Laboratory of Robotics Shenyang Institute of AutomationChinese Academy of SciencesShenyang 110169China Beijing Institute of Spacecraft System Engineering CAST Beijing 100094China

Robotic grippers have been used in industry as end-effectors but are usually limited to operations in pre-defined ***,few devices can capture irregularly shaped dynamic targets in space,underwater and other unstructured *** this paper,a novel continuum arm group mechanism inspired by the morphology and motions of sea anemones is *** is able to dissipate and absorb the kinetic energy of a fast moving target in omni-direction and utilize multiple arms to wrap and lock the target without accurate positioning ***-driven actuation systems are implemented in the individual continuum arms,achieving both bending motion and stiffness *** finite element method,the influence of different configurations of the continuum arm group on the capture performance is analyzed.A robotic prototype is constructed and tested,showing the presented arm group mechanism has high adaptability to capture targets with different sizes,shapes,and incident angles.

关键词： Non-cooperative targets Continuum arm group Wire-driven mechanism

Analysis of Pseudo-Labeling for Online Source-Free Universal Domain Adaptation

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Schlachter, Pascal Fuss, Jonathan Yang, Bin Institute of Signal Processing and System Theory University of Stuttgart Germany

A domain (distribution) shift between training and test data often hinders the real-world performance of deep neural networks, necessitating unsupervised domain adaptation (UDA) to bridge this gap. Online source-free UDA has emerged as a solution for practical scenarios where access to source data is restricted and target data is received as a continuous stream. However, the open-world nature of many real-world applications additionally introduces category shifts meaning that the source and target label spaces may differ. Online source-free universal domain adaptation (SF-UniDA) addresses this challenge. Existing methods mainly rely on self-training with pseudo-labels, yet the relationship between pseudo-labeling and adaptation outcomes has not been studied yet. To bridge this gap, we conduct a systematic analysis through controlled experiments with simulated pseudo-labeling, offering valuable insights into pseudo-labeling for online SF-UniDA. Our findings reveal a substantial gap between the current state-of-the-art and the upper bound of adaptation achieved with perfect pseudo-labeling. Moreover, we show that a contrastive loss enables effective adaptation even with moderate pseudo-label accuracy, while a cross-entropy loss, though less robust to pseudo-label errors, achieves superior results when pseudo-labeling approaches perfection. Lastly, our findings indicate that pseudo-label accuracy is in general more crucial than quantity, suggesting that prioritizing fewer but high-confidence pseudo-labels is beneficial. Overall, our study highlights the critical role of pseudo-labeling in (online) SF-UniDA and provides actionable insights to drive future advancements in the field. Our code is available at https://***/pascalschlachter/PLAnalysis. Copyright © 2025, The Authors. All rights reserved.

关键词： Deep neural networks

Estimation of bivariate structural causal models by variational Gaussian process regression under likelihoods parametrised by normalising flows

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Reick, Nico Wiewel, Felix Bartler, Alexander Yang, Bin Institute of Signal Processing and System Theory University of Stuttgart Germany

One major drawback of state-of-the-art artificial intelligence is its lack of explainability. One approach to solve the problem is taking causality into account. Causal mechanisms can be described by structural causal models. In this work, we propose a method for estimating bivariate structural causal models using a combination of normalising flows applied to density estimation and variational Gaussian process regression for post-nonlinear models. It facilitates causal discovery, i.e. distinguishing cause and effect, by either the independence of cause and residual or a likelihood ratio test. Our method which estimates post-nonlinear models can better explain a variety of real-world causeeffect pairs than a simple additive noise model. Though it remains difficult to exploit this benefit regarding all pairs from the Tübingen benchmark database, we demonstrate that combining the additive noise model approach with our method significantly enhances causal discovery. Copyright © 2021, The Authors. All rights reserved.

关键词： Additives

A Deep-Learning-Aided Pipeline for Efficient Post-Silicon Tuning

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Liao, Yiwen Yang, Bin Rivoir, Jochen Latty, Raphaël Institute of Signal Processing and System Theory University of Stuttgart Germany Applied Research and Venture Team Advantest Europe GmbH Germany

In post-silicon validation, tuning is to find the values for the tuning knobs, potentially as a function of process parameters and/or known operating conditions. In this sense, an more efficient tuning requires identifying the most critical tuning knobs and process parameters in terms of a given figure-of-merit for a Device Under Test (DUT). This is often manually conducted by experienced experts. However, with increasingly complex chips, manual inspection on a large amount of raw variables has become more challenging. In this work, we leverage neural networks to efficiently select the most relevant variables and present a corresponding deep-learning-aided pipeline for efficient tuning. Copyright © 2022, The Authors. All rights reserved.

关键词： Pipelines

Contrastive learning and self-training for unsupervised domain adaptation in semantic segmentation

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Marsden, Robert Alexander Bartler, Alexander Döbler, Mario Yang, Bin Institute of Signal Processing and System Theory University of Stuttgart Germany

Deep convolutional neural networks have considerably improved state-of-the-art results for semantic segmentation. Nevertheless, even modern architectures lack the ability to generalize well to a test dataset that originates from a different domain. To avoid the costly annotation of training data for unseen domains, unsupervised domain adaptation (UDA) attempts to provide efficient knowledge transfer from a labeled source domain to an unlabeled target domain. Previous work has mainly focused on minimizing the discrepancy between the two domains by using adversarial training or self-training. While adversarial training may fail to align the correct semantic categories as it minimizes the discrepancy between the global distributions, self-training raises the question of how to provide reliable pseudo-labels. To align the correct semantic categories across domains, we propose a contrastive learning approach that adapts category-wise centroids across domains. Furthermore, we extend our method with self-training, where we use a memory-efficient temporal ensemble to generate consistent and reliable pseudo-labels. Although both contrastive learning and self-training (CLST) through temporal ensembling enable knowledge transfer between two domains, it is their combination that leads to a symbiotic structure. We validate our approach on two domain adaptation benchmarks: GTA5 → Cityscapes and SYNTHIA → Cityscapes. Our method achieves better or comparable results than the state-of-the-art. We will make the code publicly available. Copyright © 2021, The Authors. All rights reserved.

关键词： Knowledge management