检索结果-内蒙古大学图书馆

Deep Reference Priors: What is the best way to pretrain a model?

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Gao, Yansong Ramesh, Rahul Chaudhari, Pratik Applied Mathematics and Computational Sciences University of Pennsylvania United States Computer and Information Science University of Pennsylvania United States Electrical and Systems Engineering University of Pennsylvania United States

What is the best way to exploit extra data-be it unlabeled data from the same task, or labeled data from a related task-to learn a given task? This paper formalizes the question using the theory of reference priors. Reference priors are objective, uninformative Bayesian priors that maximize the mutual information between the task and the weights of the model. Such priors enable the task to maximally affect the Bayesian posterior, e.g., reference priors depend upon the number of samples available for learning the task and for very small sample sizes, the prior puts more probability mass on low-complexity models in the hypothesis space. This paper presents the first demonstration of reference priors for medium-scale deep networks and image-based data. We develop generalizations of reference priors and demonstrate applications to two problems. First, by using unlabeled data to compute the reference prior, we develop new Bayesian semi-supervised learning methods that remain effective even with very few samples per class. Second, by using labeled data from the source task to compute the reference prior, we develop a new pretraining method for transfer learning that allows data from the target task to maximally affect the Bayesian posterior. Empirical validation of these methods is conducted on image classification datasets. Code is available at https://***/grasp-lyrl/deep_reference_priors. © 2022, CC BY.

关键词： Classification (of information)

Inverse Design with Dynamic Mode Decomposition

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Zhu, Yunpeng Cheng, Liangliang Jing, Anping Huo, Hanyu Lang, Ziqiang Zhang, Bo Kutz, J. Nathan School of Engineering and Material Science Queen Mary University of London LondonE1 4NS United Kingdom ENTEG Faculty of Science and Engineering University of Groningen Groningen9747 AG Netherlands School of Mechanical Engineering Ningxia University Yinchuan750021 China School of Chemistry and Material Science University of Science and Technology of China Hefei230026 China School of Electrical and Electronic Engineering The University of Sheffield SheffieldS10 2TN United Kingdom Department of Applied Mathematics and Electrical and Computer Engineering University of Washington SeattleWA United States

We introduce a computationally efficient method for the automation of inverse design in science and engineering. Based on simple least-square regression, the underlying dynamic mode decomposition algorithm can be used to construct a low-rank subspace spanning multiple experiments in parameter space. The proposed inverse design dynamic mode composition (ID-DMD) algorithm leverages the computed low-dimensional subspace to enable fast digital design and optimization on laptop-level computing, including the potential to prescribe the dynamics themselves. Moreover, the method is robust to noise, physically interpretable, and can provide uncertainty quantification metrics. The architecture can also efficiently scale to large-scale design problems using randomized algorithms in the ID-DMD. The simplicity of the method and its implementation are highly attractive in practice, and the ID-DMD has been demonstrated to be an order of magnitude more accurate than competing methods while simultaneously being 3-5 orders faster on challenging engineering design problems ranging from structural vibrations to fluid dynamics. Due to its speed, robustness, interpretability, and ease-of-use, ID-DMD in comparison with other leading machine learning methods represents a significant advancement in data-driven methods for inverse design and optimization, promising a paradigm shift in how to approach inverse design in *** Codes 37M05, 37M10, 37M21 Copyright © 2025, The Authors. All rights reserved.

关键词： Dynamic mode decomposition

On the relevance of lift force modelling in turbulent wall flows with small inertial particles

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Gao, Wei Shi, Pengyu Parsani, Matteo Costa, Pedro Mechanical Engineering Physical Science and Engineering Division King Abdullah University of Science and Technology Thuwal23955-6900 Saudi Arabia Applied Mathematics and Computational Science Computer Electrical and Mathematical Science and Engineering Division Extreme Computing Research Center King Abdullah University of Science and Technology Thuwal23955-6900 Saudi Arabia Université de Toulouse CNRS Toulouse France Helmholtz-Zentrum Dresden – Rossendorf Institute of Fluid Dynamics Dresden01328 Germany Process & Energy Department TU Delft Leeghwaterstraat 39 Delft2628 CB Netherlands

In particle-laden turbulent wall flows, lift forces can influence the near-wall turbulence. This has been recently observed in particle-resolved simulations, which, however, are too expensive to be used in upscaled models. Instead, point-particle simulations have been the method of choice to simulate the dynamics of these flows during the last decades. While this approach is simpler, cheaper, and physically sound for small inertial particles in turbulence, some issues remain. In the present work, we address challenges associated with lift force modelling in turbulent wall flows and the impact of lift forces in the near-wall flow. We performed direct numerical simulations (DNS) of small inertial point particles in turbulent channel flow for fixed Stokes number and mass loading while varying the particle size. Our results show that the particle dynamics in the buffer region, causing the apparent particle-to-fluid slip velocity to vanish, raises major challenges for accurately modelling lift forces. While our results confirm that lift forces have little influence on particle dynamics for sufficiently small particle sizes, for inner-scaled diameters of order one and beyond, lift forces become quite important near the wall. The different particle dynamics under lift forces results in the modulation of streamwise momentum transport in the near-wall region. We analyze this lift-induced turbulence modulation for different lift force models, and the results indicate that realistic models are critical for particle-modelled simulations to correctly predict turbulence modulation by particles in the near-wall region. © 2024, CC BY.

关键词： Wall flow

Some hybrid Riemannian conjugate gradient methods with restart strategy

学校读者我要写书评

暂无评论

Franklin Open

Franklin Open 2025年 10卷

作者： Salihu, Nasiru Kumam, Poom Salisu, Sani Khammahawong, Konrawut Center of Excellence in Theoretical and Computational Science (TaCS-CoE) Fixed Point Research Laboratory Fixed Point Theory and Applications Research Group Faculty of Science King Mongkut's University of Technology Thonburi (KMUTT) Bangkok 10140 Thailand Department of Mathematics Faculty of Sciences Modibbo Adama University Yola 652105 Nigeria KMUTT-Fixed Point Research Laboratory Room SCL 802 Science Laboratory Building Department of Mathematics Faculty of Science King Mongkut's University of Technology Thonburi (KMUTT) Bangkok 10140 Thailand Department of Mathematics Faculty of Natural and Applied Sciences Sule Lamido University Kafin Hausa Jigawa Nigeria Applied Mathematics for Science and Engineering Research Unit (AMSERU) Department of Mathematics and Computer Science Faculty of Science and Technology Rajamangala University of Technology Thanyaburi (RMUTT) Pathum Thani 12110 Thailand

Hybrid nonlinear conjugate gradient (CG) methods are known to be efficient and have less memory requirement for solving optimization problems in Euclidean spaces. The formulations strategically switch between two or more CG parameters to enhance their general structures. In this work, some hybrid Riemannian CG methods are presented, which are sufficiently descent irrespective of the line search technique employed and converges globally using scaled version of the nonexpensive condition. Finally, the numerical experiments are performed to confirm the strength of the schemes by considering some Riemannian optimization problems. The numerical results obtained when compared to some classical Riemannian CG methods indicated the efficiency of the proposed hybrid Riemannian CG methods. © 2025 The Authors

关键词： Global convergence Retraction Riemannian conjugate gradient method Riemannian optimization Riemannian wolfe conditions Vector transport

Convective diffusive thermal flow over an inclined surface with viscous dissipation and aligned magnetic field applications

学校读者我要写书评

暂无评论

Partial Differential Equations in applied mathematics 2024年 12卷

作者： Mopuri, Obulesu Al-Khaled, Kamel Govindan, Vediyappan Ganteda, Charankumar Ganjikunta, Aruna Abdullaeva, Barno Ahmad, Furqan Khan, Sami Ullah Waqas, M. Piriadarshani, D. Department of Mathematics Siddharth Institute of Engineering & Technology (Autonomous) A.P. Puttur 517583 India Department of Mathematics & Statistics Jordan University of Science and Technology P.O. Box 3030 Irbid 22110 Jordan Department of Mathematics Hindustan Institute of Technology and Science (Deemed to be University) Padur Chennai India Department of Engineering Mathematics College of Engineering Koneru Lakshmaiah Education Foundation Vaddeswaram India Department of Mathematics GITAM University Telangana Hyderabad 502329 India Department of Mathematics and Information Technologies Vice-Rector for Scientific Affairs Tashkent State Pedagogical University Tashkent Uzbekistan Western Caspian University Baku Azerbaijan Department of Mathematics Namal University Mianwali 42250 Pakistan NUTECH School of Applied Science and Humanities National University of Technology Islamabad 44000 Pakistan Department of Computer Science and Mathematics Lebanese American University Beirut Lebanon

This investigation incorporating the fluctuation in heat and mass transfer associated to the mixed convection magnetized flow of viscous fluid due to inclined surface with porous media. The contribution of Soret effects and viscous dissipation appliances is addressed. Furthermore, the heat transfer improvement is also assessed by thermal radiation, heat source and joule heating effects. The chemical reaction enrollment is also studied for concentration phenomenon. The convection of problem into non-dimensional framework is based on implication of new variables. The perturbation technique is followed to tracking the analytical outcomes. Physical visualization and interpretation of results under the influence of perturbed parameters have been studied. It is observed that heat and mass transfer enhances due to Soret number. Presence of chemical reaction leads to decrement of concentration profile. Claimed results presents applications in heat and mass transfer processes, chemical reaction, manufacturing systems, chemical engineering, extrusion processes etc. © 2024 The Author(s)

关键词： Heat and mass transfer Joule heating Mixed convection Perturbation simulation Viscous dissipation

The Impact of Selfish Behavior in Load Balancing Games

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Bilò, Vittorio Vinci, Cosimo Department of Mathematics and Physics "Ennio De Giorgi" University of Salento Lecce73100 Italy Department of Computer Engineering Electrical Engineering and Applied Mathematics University of Salerno Fisciano84084 Italy Gran Sasso Science Institute L’Aquila67100 Italy

To what extent does the structure of the players’ strategy space influence the efficiency of decentralized solutions in congestion games? In this work, we investigate whether better performance are possible when restricting to load balancing games in which players can only choose among single resources. We consider three different solutions concepts, namely, approximate pure Nash equilibria, approximate one-round walks generated by selfish players aiming at minimizing their personal cost and approximate one-round walks generated by cooperative players aiming at minimizing the marginal increase in the sum of the players’ personal costs. The last two concepts can be interpreted as solutions of greedy online algorithms for the related resource selection problem. We show that, under fairly general latency functions on the resources, better bounds cannot be achieved if players are either weighted or asymmetric. On the positive side, we prove that, under mild assumptions on the latency functions, improvements on the performance of approximate pure Nash equilibria are possible for load balancing games with weighted and symmetric players in the case of identical resources. We also design lower bounds on the performance of one-round walks in load balancing games with unweighted players and identical *** Codes 91A99 (Primary) 68W27, 90C05 (Secondary) © 2022, CC BY.

关键词： Resource allocation

A deep Bi-CapsNet for analysing ECG signals to classify cardiac arrhythmia

学校读者我要写书评

暂无评论

computers in Biology and Medicine 2025年 189卷 109924-109924页

作者： Anitha, T. Aanjankumar, S. Dhanaraj, Rajesh Kumar Pamucar, Dragan Simic, Vladimir Department of Computer Science and Engineering Sathyabama Institute of Science and Technology Chennai India VIT Bhopal University Bhopal- Indore Highway Kothrikalan Madhya Pradesh Sehore466114 India Pune India Department of Operations Research and Statistics Faculty of Organizational Sciences University of Belgrade Belgrade Serbia Department of Applied Mathematical Science College of Science and Technology Korea University Sejong30019 Korea Republic of Department of Mechanics and Mathematics Western Caspian University Baku Azerbaijan University of Belgrade Faculty of Transport and Traffic Engineering Vojvode Stepe 305 Belgrade11010 Serbia Yuan Ze University College of Engineering Department of Industrial Engineering and Management Taoyuan City320315 Taiwan

— In recent times, the electrocardiogram (ECG) has been considered as a significant and effective screening mode in clinical practice to assess cardiac arrhythmias. Precise feature extraction and classification are considered as essential concerns in the automated prediction of heart disease. A deep bi-directional capsule network (Bi-CapsNet) uses a new method based on an intelligent deep learning (DL) classifier model to make the classification process very accurate. Initially, the input ECG signal data are acquired and the preprocessing steps such as DC drift, normalization, LPF filtering, spectrogram analysis, and artifact removal are applied. After preprocessing the data, the Deep Ensemble CNN-RNN approach is employed for feature extraction. Finally, the Deep Bi-CapsNet model is used to predict and classify the cardiac arrhythmia. For performance validation, the dataset is referred to the MIT-BIH arrhythmia database, which selects five different types of arrhythmias from the ECG waveform to estimate the proposed model. Various ECG arrhythmia categories, including Normal (NOR), Right Bundle Branch Block (RBBB), Premature Ventricular Contraction (PVC), Atrial Premature Beat (APB), and Left Bundle Branch Block (LBBB) have been identified. For performance analysis, the metrics such as precision, accuracy, F1-score, error rate, sensitivity, false positive rate, specificity, Mathew coefficient, Kappa coefficient, and outcomes are included and compared with the traditional methods to validate the effectiveness of the implemented scheme. The proposed scheme has achieved an overall accuracy rate of approximately 97.19 % compared to the traditional deep learning models like CNN (89.87 %), FTBO (85 %), and Capsule Network (97.0 %). The comparison results indicate that the proposed hybrid model outperforms these traditional models. © 2025 Elsevier Ltd

关键词： Electrocardiograms

A Three-Level Extension for Fast and Robust Overlapping Schwarz (FROSch) Preconditioners with Reduced Dimensional Coarse Space 26th

学校读者我要写书评

暂无评论

A Three-Level Extension for Fast and Robust Overlapping Schw...

26th International Conference on Domain Decomposition Methods, 2020

作者： Heinlein, Alexander Klawonn, Axel Rheinbach, Oliver Röver, Friederike Faculty of Electrical Engineering Mathematics & Computer Science Delft Institute of Applied Mathematics Delft University of Technology Mekelweg 4 CD Delft72628 Netherlands Department Mathematik/Informatik Universität zu Köln Weyertal 86-90 Köln50923 Germany Center for Data and Simulation Science University of Cologne Cologne Germany Institut für Numerische Mathematik und Optimierung Technische Universität Bergakademie Freiberg Akademiestr. 6 Freiberg09599 Germany

The Fast and Robust Overlapping Schwarz framework [7, 8], which is part of the Trilinos Software library [18], contains a parallel implementation of the generalized Dryja–Smith–Widlund (GDSW) preconditioner. © ...

ISBN: (纸本)9783030950248

关键词：

Automated Vision-Based System for Real-Time Detection and Classification of Rice Diseases using YOLO-NAS

学校读者我要写书评

暂无评论

Automated Vision-Based System for Real-Time Detection and Cl...

Trends in engineering Systems and Technologies (ICTEST), International Conference on

作者： Jose P Rekha S Ponkavin B Hemkiran S Sheeja R P.V. Gopirajan Department of Computer Science and Engineering Vel Tech Rangarajan Dr. Sagunthala R&D Institute of Science and Technology Chennai Tamil Nadu Department of Mathematics DMI Engineering College Aralvoimozhi Tamil Nadu Department of Mechanical Engineering Rohini College of Engineering Palkulam Tamil Nadu Department of CSE PSG Institute of Technology and Applied Research Coimbatore Tamil Nadu Department of AI&DS R.M.K. Engineering College Chennai Tamil Nadu Department of Computational Intelligence SRM Institute of Science and Technology Kattankulathur Tamil Nadu

ISBN: (数字)9798331505370

ISBN: (纸本)9798331505387

This proposed work presents an innovative approach to agricultural disease management through the integration of the YOLO-NAS (You Only Look Once Neural Architecture Search) algorithm for real-time rice leaf disease detection. Focusing on five major diseases—brown spot, bacterial leaf blight, hispa, dead heart, and bacterial panicle blight—the system leverages deep learning and computer vision to offer rapid and accurate disease identification. By providing early detection insights, the tool aids farmers in making informed decisions, enhancing precision agriculture, and fostering sustainable farming practices. The user-friendly interface ensures widespread adoption, empowering farmers to monitor disease progression and take timely action. This approach mitigates crop losses, improves yield, and demonstrates the transformative potential of AI in agriculture and the result F1_Score, Precision, Accuracy and Map/ap scores as 76%,90%,67% respectively. The study highlights the importance of technological innovation in addressing critical agricultural challenges and advancing global food security.

关键词： YOLO Deep learning Technological innovation Microorganisms Food security Crops Real-time systems Neural architecture search Diseases Farming