检索结果-内蒙古大学图书馆

Physics-Driven Deep Learning for computational Magnetic Resonance Imaging: Combining physics and machine learning for improved medical imaging

引用

IEEE SIGNAL PROCESSING MAGAZINE 2023年第1期40卷 98-114页

作者： Hammernik, Kerstin Kuestner, Thomas Yaman, Burhaneddin Huang, Zhengnan Rueckert, Daniel Knoll, Florian Akcakaya, Mehmet Imperial Coll London Dept Comp London SW7 2AZ England Univ Hosp Tubingen D-72076 Tubingen Germany Univ Minnesota Elect Engn Minneapolis MN 55455 USA NYU Sch Med Ctr Biomed Imaging New York NY 10016 USA Friedrich Alexander Univ Erlangen Nuremberg Dept Artificial Intelligence Biomed Engn D-91052 Erlangen Germany Univ Minnesota Minneapolis MN 55455 USA

Physics-driven deep learning methods have emerged as a powerful tool for computational magnetic resonance imaging (MRI) problems, pushing reconstruction performance to new limits. This article provides an overview of the recent developments in incorporating physics information into learning-based MRI reconstruction. We consider inverse problems with both linear and nonlinear forward models for computational MRI and review the classical approaches for solving these. We then focus on physics-driven deep learning approaches, covering physics-driven loss functions, plug-and-play (PnP) methods, generative models, and unrolled networks. We highlight domain-specific challenges, such as real- and complex-valued building blocks of neural networks, and translational applications in MRI with linear and nonlinear forward models. Finally, we discuss common issues and open challenges, and we draw connections to the importance of physics-driven learning when combined with other downstream tasks in the medical imaging pipeline.

关键词： Deep learning Inverse problems Magnetic resonance imaging computational modeling Pipelines Signal processing Task analysis

来源：评论

学校读者我要写书评

暂无评论

Fold3D: Rethinking and Parallelizing computational and Communicational Tasks in the Training of Large DNN Models

引用

IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS 2023年第5期34卷 1432-1449页

作者： Li, Fanxin Zhao, Shixiong Qing, Yuhao Chen, Xusheng Guan, Xiuxian Wang, Sen Zhang, Gong Cui, Heming Univ Hong Kong Dept Comp Sci Hong Kong Peoples R China Huawei Tech Co Ltd Theory Lab Labs 2012 Hong Kong Peoples R China

Training a large DNN (e.g., GPT3) efficiently on commodity clouds is challenging even with the latest 3D parallel training systems (e.g., Megatron v3.0). In particular, along the pipeline parallelism dimension, computational tasks that produce a whole DNN's gradients with multiple input batches should be concurrently activated;along the data parallelism dimension, a set of heavy-weight communications (for aggregating the accumulated outputs of computational tasks) is inevitably serialized after the pipelined tasks, undermining the training performance (e.g., in Megatron, data parallelism caused all GPUs idle for over 44% of the training time) over commodity cloud networks. To deserialize these communicational and computational tasks, we propose the AIAO scheduling (for 3D parallelism) which slices a DNN into multiple segments, so that the computational tasks processing the same DNN segment can be scheduled together, and the communicational tasks that synchronize this segment can be launched and overlapped (deserialized) with other segments' computational tasks. We realized this idea in our Fold3D training system. Extensive evaluation shows Fold3D eliminated most of the all-GPU 44% idle time in Megatron (caused by data parallelism), leading to 25.2%-42.1% training throughput improvement compared to four notable baselines over various settings;Fold3D's high performance scaled to many GPUs.

关键词： Task analysis Training Three-dimensional displays Processor scheduling Graphics processing units Pipeline processing computational modeling 3D parallelism DNN deep learning distributed training GPU machine learning pipeline parallelism

来源：评论

学校读者我要写书评

暂无评论

Deep Optical Coding Design in computational Imaging: A data-driven framework

引用

IEEE SIGNAL PROCESSING MAGAZINE 2023年第2期40卷 75-88页

作者： Arguello, Henry Bacca, Jorge Kariyawasam, Hasindu Vargas, Edwin Marquez, Miguel Hettiarachchi, Ramith Garcia, Hans Herath, Kithmini Haputhanthri, Udith Ahluwalia, Balpreet Singh So, Peter Wadduwage, Dushan N. Edussooriya, Chamira U. S. Univ Ind Santander Syst Engn Dept Bucaramanga 680002 Colombia UIS Bucaramanga 680002 Colombia Harvard Univ Ctr Adv Imaging Cambridge MA 02138 USA Univ Ind Santander Bucaramanga 680002 Colombia Harvard Univ Div Sci Fac Arts & Sci Cambridge MA 02138 USA UiT Arctic Univ Norway Dept Phys & Technol Tromso Norway MIT Dept Mech & Biol Engn Cambridge MA 02139 USA Harvard Univ Ctr Adv Imaging Wadduwage Lab Cambridge MA 02138 USA Univ Moratuwa Dept Elect & Telecommun Engn Moratuwa Sri Lanka

computational optical imaging (COI) systems leverage optical coding elements (CEs) in their setups to encode a high-dimensional scene in a single or in multiple snapshots and decode it by using computational algorithms. The performance of COI systems highly depends on the design of its main components: the CE pattern and the computational method used to perform a given task. Conventional approaches rely on random patterns or analytical designs to set distribution of the CE. However, the available data and algorithm capabilities of deep neural networks (DNNs) have opened a new horizon in CE data-driven designs that jointly consider the optical and computational decoders. Specifically, by modeling the COI measurements through a fully differentiable image-formation model that considers the physics-based propagation of light and its interaction with the CEs, the parameters that define the CE and the computational decoder can be optimized in an end-to-end (E2E) manner. Moreover, by optimizing just CEs in the same framework, inference tasks can be performed from pure optics. This work surveys the recent advances in CE data-driven design and provides guidelines on how to parameterize different optical elements to include them in the E2E framework. As the E2E framework can handle different inference applications by changing the loss function and the DNN, we present low-level tasks such as spectral imaging reconstruction or high-level tasks such as pose estimation with privacy preservation enhanced by using optimal task-based optical architectures. Finally, we illustrate classification and 3D object-recognition applications performed at the speed of the light using all-optics DNNs.

关键词： Privacy Three-dimensional displays Optical design computational modeling Signal processing algorithms Optical imaging Encoding

来源：评论

学校读者我要写书评

暂无评论

A Systematic Literature Review in Robotics Experiential Learning With computational and Adversarial Thinking

引用

IEEE ACCESS 2023年 11卷 44806-44827页

作者： Adnan, Noridayu Abdullah, Siti Norul Huda Sheikh Raja Yusof, Raja Jamilah Zainal, Noor Faridatul Ainun Qamar, Faizan Yadegaridehkordi, Elaheh Univ Kebangsaan Malaysia Fac Informat Sci & Technol Bangi Malaysia Univ Malaya Fac Comp Sci & Informat Technol Dept Software Engn Kuala Lumpur Malaysia

The rise of the Industrial Revolution 4.0 and the increasing reliance on the digital economy drive the need for a new set of skills, especially in robotics learning, that includes computational thinking (CT) and adversarial thinking (AT) for the young generation. The need for CT-related skills includes various fields, such as robotics, engineering, computer science, mathematics, music, arts, and humanities. Therefore, adopting robotic learning with CT and AT can enhance learning skills over the conventional learning model. This paper presents a systematic literature review on CT and AT practices in robotics learning to improve educational methods. This study conducts a systematic literature review from four databases: ACM, Scopus, IEEE Xplore, and ScienceDirect. Sixty-five studies in robotics learning to increase CT and AT skills were analyzed by applying the inclusion and exclusion criteria. The study's findings show that CT and AT are significant in training students to engage in robotics learning activities. These considerations will lead to strengthening their skill and critical thinking. The study also suggests that integrating these skills can prepare teachers for critical thinking and boost student learning. The findings suggest that CT and AT can directly adopt digital adversarial learning skills to improve overall robotics learning activities. For future studies, the difference in learning ages related to robotics activities with CT and AT applications can be studied to deeply comprehend the effectiveness of CT and AT applications.

关键词： Robots computational modeling Systematics Education Mathematical models Learning systems Adaptation models Fourth Industrial Revolution Robotics computational thinking adversarial thinking experiential learning model

来源：评论

学校读者我要写书评

暂无评论

An Application of a Higher computational Efficiency Method in Thermo-Hydraulic Calculations of First Wall in Fusion Blanket

引用

IEEE TRANSACTIONS ON PLASMA SCIENCE 2023年第12期51卷 3688-3695页

作者： Guo, Likai Wu, Xinghua Zhao, Shengqiang Huo, Minhan Southwestern Inst Phys Chengdu 610041 Sichuan Peoples R China Tsinghua Univ Beijing 100190 Peoples R China

In the field of computer science, computational efficiency is an important indicator for measuring computational methods. The internal structure of the fusion reactor blanket is complex and has a large spatial scale, making the computational cost of the overall thermal-hydraulic calculation particularly large. Therefore, it is necessary to optimize the computational efficiency. This article introduces a method to improve the computational efficiency of thermo-hydraulic calculations and presents a preliminary validation in the context of the first wall of a blanket. By contrasting different fitting methods for the thermophysical parameters of helium coolant in a helium-cooled solid breeder blanket, it is observed that utilizing a power function to fit helium density, with a similar fitting quality, leads to a certain degree of improvement in computational efficiency compared to commonly used multiple polynomial fitting, in the same order of magnitude. Subsequently, a method employing matrix correlation analysis for dimensionality reduction is proposed to increase the computational efficiency. Notably, there is a significant optimization in computational efficiency after applying dimensionality reduction. Furthermore, the proposed method is validated using User-Defined Functions (UDFs) within the commercial software FLUENT. Finally, by comparing the results of calculations using the power function density formula and matrix dimensionality reduction with theoretical hydraulic calculations, the optimized computational approach is found to yield reliable results.

关键词： computational efficiency Fitting Helium Correlation Mathematical models computational modeling Dimensionality reduction fusion reactor matrix correlation analysis thermal-hydraulics user-defined function (UDF)

来源：评论

学校读者我要写书评

暂无评论

An Uncertainty Analysis on Finite Difference Time-Domain Computations With Artificial Neural Networks: Improving Accuracy While Maintaining Low computational Costs.

引用

IEEE ANTENNAS AND PROPAGATION MAGAZINE 2023年第1期65卷 60-70页

作者： Hu, Runze Monebhurrun, Vikass Himeno, Ryutaro Yokota, Hideo Costen, Fumie Univ Manchester Manchester M13 9PL Lancs England Tsinghua Univ Beijing 100084 Peoples R China CentraleSupelec F-91192 Gif Sur Yvette France Juntendo Univ Tokyo 1138421 Japan RIKEN Ctr Adv Photon Wako Saitama Japan Univ Rytkus Fac Med Nishihara Okinawa 9030213 Japan

Artificial neural networks (ANNs) have appeared as a potential alternative for uncertainty quantification (UQ) in the finite difference time-domain (FDTD) computation. They are applied to build a surrogate model for the computation-intensive FDTD simulation and to bypass the numerous simulations required for UQ. However, when the surrogate model utilizes an ANN, a considerable number of data are generally required for high accuracy, and generating such large quantities of data becomes computationally prohibitive. To address this drawback, a number of adaptations for ANNs are proposed, which additionally improves the accuracy of ANNs in UQ for the FDTD computation while maintaining a low computational cost. The proposed algorithm is tested for application in bioelectromagnetics, and considerable speed up, as well as the improved accuracy of UQ, is observed compared to traditional methods such as the nonintrusive polynomial chaos (NIPC) method.

关键词： Finite difference methods Time-domain analysis computational modeling Uncertainty Biological system modeling Neurons Data models

来源：评论

学校读者我要写书评

暂无评论

Enhanced Local Stabilization of Constrained N-TS Fuzzy Systems With Lighter computational Burden

引用

IEEE TRANSACTIONS ON FUZZY SYSTEMS 2023年第3期31卷 1064-1070页

作者： Xie, Xiangpeng Yang, Fuyi Wan, Lei Xia, Jianwei Shi, Kaibo Nanjing Univ Posts & Telecommun Inst Adv Technol Nanjing 210023 Peoples R China Liaocheng Univ Sch Math Sci Liaocheng 252000 Peoples R China Chengdu Univ Sch Informat Sci & Engn Chengdu 610106 Peoples R China

The constrained Takagi-Sugeno fuzzy systems with local nonlinear models (N-TS fuzzy systems) are much more practical than the conventional T-S fuzzy models, but their applications are seriously affected by either the great conservatism or the heavy computational burden. In this short article, an enhanced local stabilization of constrained N-TS fuzzy systems is investigated with the purpose of not only reducing conservatism but also relieving computational burden. To do this, a new class of threshold functions are established in order to achieve real-time online assessment of the variation of the normalized fuzzy weighting functions (NFWFs) at different sampling instants, i.e., the one-step-past one and the current one. Thus, a key alternative is proposed for eliminating all the involved terms containing the one-step-past NFWFs while their main information is retained. Since our obtained design criteria are merely dependent on the current NFWFs, the conservatism can be reduced at the expense of lighter computational burden. The benefits of our proposed results are validated by implementing several objective comparisons with those recent results.

关键词： Fuzzy systems Linear matrix inequalities Lyapunov methods Costs computational modeling Closed loop systems Telecommunications Constrained systems fuzzy control fuzzy stabilization input saturation

来源：评论

学校读者我要写书评

暂无评论

Long-Horizon Robust Direct Model Predictive Control for Medium-Voltage Induction Motor Drives With Reduced computational Complexity

引用

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS 2023年第2期59卷 1775-1787页

作者： Tregubov, Andrei Karamanakos, Petros Ortombina, Ludovico Tampere Univ Fac Informat Technol & Commun Sci Tampere Finland Univ Padua Dept Ind Engn I-35122 Padua Italy

This paper proposes a long-horizon direct model predictive control (MPC) with reference tracking for medium-voltage (MV) drives that achieves favorable steady-state and transient behavior. However, as MPC is a model-based method, it is susceptible to parameter mismatches and variations of the machine. Moreover, even though a long prediction horizon significantly improves the steady-state behavior of the drive, it significantly increases the computational complexity of the direct MPC problem, rendering its real-time implementation a challenging-if not impossible-task. Motivated by these shortcomings of long-horizon direct MPC, this paper also aims to address them by enhancing the robustness of the developed control strategy, while keeping its computational complexity modest. To achieve the former, a prediction model suitable for MV drive systems is adopted that facilitates the effective estimation of the total leakage inductance of the machine. For the latter, the objective function of the MPC problem is formulated such that, even though the drive behavior is computed over a long prediction interval, only a few changes in the candidate switch positions are considered. The effectiveness of the proposed modeling, control, and estimation approaches is validated with hardware-in-the-loop (HIL) tests for an MV drive consisting of a three-level neutral point clamped (NPC) inverter and an induction machine (IM).

关键词： Stators computational modeling Rotors Predictive models computational complexity Real-time systems Estimation AC drives direct control hardware-in-the-loop (HIL) simulations medium-voltage (MV) drives model predictive control (MPC) robust control

来源：评论

学校读者我要写书评

暂无评论

Tacchi: A Pluggable and Low computational Cost Elastomer Deformation Simulator for Optical Tactile Sensors

引用

IEEE ROBOTICS AND AUTOMATION LETTERS 2023年第3期8卷 1239-1246页

作者： Chen, Zixi Zhang, Shixin Luo, Shan Sun, Fuchun Fang, Bin Kings Coll London Dept Engn London WC2 R2LS England Tsinghua Univ Inst Artificial Intelligence Beijing Natl Res Ctr Informat Sci & Technol Dept Comp Sci & Technol Beijing 100084 Peoples R China China Univ Geosci Beijing Sch Engn & Technol Beijing 100083 Peoples R China

Simulation is widely applied in robotics research to save time and resources. There have been several works to simulate optical tactile sensors that leverage either a smoothing method or Finite Element Method (FEM). However, elastomer deformation physics is not considered in the former method, whereas the latter requires a massive amount of computational resources like a computer cluster. In this work, we propose a pluggable and low computational cost simulator using the Taichi programming language for simulating optical tactile sensors, named as Tacchi. It reconstructs elastomer deformation using particles, which allows deformed elastomer surfaces to be rendered into tactile images and reveals contact information without suffering from high computational costs. Tacchi facilitates creating realistic tactile images in simulation, e.g., ones that capture wear-and-tear defects on object surfaces. In addition, the proposed Tacchi can be integrated with robotics simulators for a robot system simulation. Experiment results showed that Tacchi can produce images with better similarity to real images and achieved higher Sim2Real accuracy compared to the existing methods. Moreover, it can be connected with MuJoCo and Gazebo with only the requirement of 1G memory space in GPU compared to a computer cluster applied for FEM. With Tacchi, physical robot simulation with optical tactile sensors becomes possible.

关键词： Robot sensing systems Robots Strain computational modeling Optical sensors Optical imaging Tactile sensors Force and tactile sensing robot manipulation simulation of tactile sensors

来源：评论

学校读者我要写书评

暂无评论

A computational Model of Coping and Decision Making in High-Stress, Uncertain Situations: An Application to Hurricane Evacuation Decisions

引用

IEEE TRANSACTIONS ON AFFECTIVE COMPUTING 2023年第3期14卷 2539-2556页

作者： Yongsatianchot, Nutchanon Marsella, Stacy Northeastern Univ Khoury Coll Comp Sci Boston MA 02115 USA

People often encounter highly stressful, emotion-evoking situations. modeling and predicting people's behavior in such situations, how they cope, is a critical research topic. To that end, we propose a computational model of coping that casts Lazarus's theory of coping into a Partially Observable Markov Decision Process (POMDP) framework. This includes an appraisal process that models the factors leading to stress by assessing a person's relation to the environment and a coping process that models how people seek to reduce stress by directly altering the environment or changing one's beliefs and goals. We evaluated the model's assumptions in the context of a high-stress situation, hurricanes. We collected questionnaire data from major U.S. hurricanes in 2018 to evaluate the model's features for appraisal calculation. We also conducted a series of controlled experiments simulating a hurricane experience to investigate how people change their beliefs and goals to cope with the situation. The results support the model's assumptions showing that the proposed features are significantly associated with the evacuation decisions and people change their beliefs and goals to cope with the situation.

关键词： Hurricanes Appraisal computational modeling Regulation Psychology Predictive models Decision making computational model stress hurricane coping appraisal

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：