检索结果-内蒙古大学图书馆

Accumulation bit-width scaling for ultra-low precision training of deep networks

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Sakr, Charbel Wang, Naigang Chen, Chia-Yu Choi, Jungwook Agrawal, Ankur Shanbhag, Naresh Gopalakrishnan, Kailash Dept. of Electrical and Computer Engineering University of Illinois at Urbana-Champaign IBM T.J. Watson Research Center

Efforts to reduce the numerical precision of computations in deep learning training have yielded systems that aggressively quantize weights and activations, yet employ wide high-precision accumulators for partial sums in inner-product operations to preserve the quality of convergence. The absence of any framework to analyze the precision requirements of partial sum accumulations results in conservative design choices. This imposes an upper-bound on the reduction of complexity of multiply-accumulate units. We present a statistical approach to analyze the impact of reduced accumulation precision on deep learning training. Observing that a bad choice for accumulation precision results in loss of information that manifests itself as a reduction in variance in an ensemble of partial sums, we derive a set of equations that relate this variance to the length of accumulation and the minimum number of bits needed for accumulation. We apply our analysis to three benchmark networks: CIFAR-10 ResNet 32, ImageNet ResNet 18 and ImageNet AlexNet. In each case, with accumulation precision set in accordance with our proposed equations, the networks successfully converge to the single precision floating-point baseline. We also show that reducing accumulation precision further degrades the quality of the trained network, proving that our equations produce tight bounds. Overall this analysis enables precise tailoring of computation hardware to the application, yielding area- and power-optimal systems. Copyright © 2019, The Authors. All rights reserved.

关键词： Digital arithmetic

Cross-sectional and longitudinal Biomarker extraction and analysis for multicentre FLAIR brain MRI

学校读者我要写书评

暂无评论

Neuroimage: Reports 2022年第2期2卷 100091-100091页

作者： DiGregorio, J. Gibicar, A. Khosravani, H. Maralani, P. Jabehdar Tardif, J.-C. Tyrrell, P.N. Moody, A.R. Khademi, A. Electrical Computer and Biomedical Engineering Dept. Ryerson University Toronto ON Canada Division of Neurology Department of Medicine University of Toronto Toronto ON Canada Department of Medical Imaging University of Toronto Toronto ON Canada Montreal Heart Institute Montreal QU Canada Department of Medicine Université de Montréal QU Canada Department of Statistical Science University of Toronto Toronto ON Canada Keenan Research Center for Biomedical Science St. Michael's Hospital Unity Health Network Toronto ON Canada Institute for Biomedical Engineering Science and Technology (iBEST) A Partnership Between St. Michael's Hospital and Ryerson University Canada

Fluid-attenuated inversion recovery (FLAIR) MRI has emerged as an important sequence for the analysis of cerebrovascular (CVD) and Alzheimer's disease (AD). Large-scale, automated cross-sectional and longitudinal cerebral biomarker extraction from FLAIR datasets could progress disease characterization, improve disease monitoring, and help to determine optimal intervention times. Despite this, most automated biomarker extraction algorithms are designed for T1-weighted or multi-modal inputs. In this work, automated tools were used to extract biomarkers from large, FLAIR-only datasets to evaluate the feasibility of this sequence to characterize healthy, AD, and CVD subjects in a similar manner to traditional approaches. Total brain volume (TBV), cerebrospinal fluid (CSF) volume, and white matter lesion (WML) volume was measured for the cross-sectional biomarkers and the corresponding annual rates of change over multiple scans represented the longitudinal biomarkers. Biomarkers were extracted from two dementia cohorts (4356 vol, 162 233 images) and one vascular disease cohort (869 vol, 42 850 images) using deep learning-based segmentation algorithms designed specifically for FLAIR. Biomarkers from all cohorts were summarized using descriptive statistics, correlation analysis, and ANCOVA to assess differences in diagnostic labels while accounting for demographic and acquisition factors. Biomarkers from FLAIR MRI had similar trends with those extracted from traditional modalities in the literature for characterizing healthy, AD, and CVD subjects. This demonstrates that FLAIR MRI can be used for end-to-end analysis of large AD and CVD datasets, which can lower acquisition costs, simplify clinical translation, and reduce measurement error associated with multi-modal approaches. © 2022 The Author(s)

关键词： Alzheimer's disease Cerebral biomarkers Cerebrovascular disease FLAIR MRI

LeakDB: A benchmark dataset for leakage diagnosis in water distribution networks 1

学校读者我要写书评

暂无评论

LeakDB: A benchmark dataset for leakage diagnosis in water d...

1st International Joint Conference in Water Distribution Systems Analysis and Computing and Control in the Water Industry, WDSA/CCWI 2018

作者： Vrachimis, Stelios G. Kyriakou, Marios S. Eliades, Demetrios G. Polycarpou, Marios M. Kios Research and Innovation Center of Excellence Dept. of Electrical and Computer Engineering University of Cyprus Nicosia Cyprus

The increase of streaming data from water utilities is enabling the development of real-Time anomaly and fault detection algorithms that can detect events, such as pipe bursts and leakages. Currently, there is not a widely accessible dataset of real or realistic leakage scenarios, which could be used as a common benchmark to compare different algorithms, as well as to support research reproducibility. In this work we propose the design of a realistic leakage dataset, the Leakage Diagnosis Benchmark (LeakDB). The dataset is comprised of a large number of artificially created but realistic leakage scenarios, on different water distribution networks, under varying conditions. Additionally, a scoring algorithm was developed in MATLAB to evaluate the results of different algorithms using various metrics. The usage of the LeakDB dataset, is demonstrated by scoring four detection algorithms. The dataset is stored on an open research data repository, and will be updated in the future with new simulation scenarios. The source code of the toolkit that generates the leakage benchmark dataset, as well as the detection algorithms used, are released as open source. © 2018 1st International WDSA / CCWI 2018 Joint Conference. All rights reserved.

关键词： MATLAB

Telecom-band hyperentangled photon pairs from a fiber-based source

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Chen, Changjia Xu, Calvin Riazi, Arash Zhu, Eric Y. Greenwood, Alexander Gladyshev, Alexey V. Kazansky, Peter G. Kirby, Brian T. Qian, Li Dept. of Electrical and Computer Engineering University of Toronto 10 King's College Rd. TorontoM5S 3G4 Canada Fiber Optics Research Center Russian Academy of Sciences 38 Vavilov Street Moscow119333 Russia Optoelectronics Research Centre University of Southampton SouthamptonSO17 1BJ United Kingdom Tulane University New OrleansLA70118 United States United States Army Research Laboratory AdelphiMD20783 United States

Hyperentanglement, the simultaneous and independent entanglement of quantum particles in multiple degrees of freedom, is a powerful resource that can be harnessed for efficient quantum information processing. In photonic systems, the two degrees of freedom (DoF) often used to carry quantum and classical information are polarization and frequency, thanks to their robustness in transmission, both in free space and in optical fibers. Telecom-band hyperentangled photons generated in optical fibers are of particular interest because they are compatible with existing fiber-optic infrastructure, and can be distributed over fiber networks with minimal loss. Here, we experimentally demonstrate the generation of telecom-band biphotons hyperentangled in both the polarization and frequency DoFs using a periodically-poled silica fiber and observe entanglement concurrences above 0.95 for both polarization and frequency DOFs. Furthermore, by concatenating a Hong-Ou-Mandel interference test for frequency entanglement and full state tomography for polarization entanglement in a single experiment, we can demonstrate simultaneous entanglement in both the polarization and frequency DOFs. The states produced by our hyperentanglement source can enable protocols such as dense coding and high-dimensional quantum key distribution. © 2021, CC BY.

关键词： Polarization

Impact of DC bias on weak optical-field-driven electron emission in nano-vacuum-gap detectors

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Turchetti, Marco Bionta, Mina R. Yang, Yujia Ritzkowsky, Felix Candido, Denis Ricardo Flatté, Michael Berggren, Karl K. Keathley, Phillip D. Research Laboratory of Electronics Massachusetts Institute of Technology CambridgeMA02139 United States Center for Free-Electron Laser Science Notkestraße 85 Hamburg22607 Germany Dept. Electrical and Computer Engineering University of Iowa Iowa CityIA52242 United States

In this work, we investigate multiphoton and optical-field tunneling emission from metallic surfaces with nanoscale vacuum gaps. Using time-dependent Schrödinger equation (TDSE) simulations, we find that the properties of the emitted photocurrent in such systems can be greatly altered by the application of only a few-volt DC bias. We find that when coupled with expected plasmonic enhancements within the nanometer-scale metallic gaps, the application of this DC bias significantly reduces the threshold for the transition to optical-field-driven tunneling from the metal surface, and could sufficiently enhance the emitted photocurrents, to make it feasible to electronically tag fJ ultrafast pulses at room temperature. Given the petahertz-scale instantaneous response of the photocurrents, and the low effective capacitance of thin-film nanoantenna devices that enables ¡ 1 fs response time, detectors that exploit this bias-enhanced surface emission from nanoscale vacuum gaps could prove to be useful for communication, petahertz electronics, and ultrafast optical-field-resolved metrology. Copyright © 2020, The Authors. All rights reserved.

关键词： Electron emission

AI Models Close to your Chest: Robust Federated Learning Strategies for Multi-site CT

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Lee, Edward H. Kelly, Brendan Altinmakas, Emre Dogan, Hakan Mohammadzadeh, Maryam Colak, Errol Fu, Steve Choudhury, Olivia Ratan, Ujjwal Kitamura, Felipe Chaves, Hernan Zheng, Jimmy Said, Mourad Reis, Eduardo Lim, Jaekwang Yokoo, Patrícia Mitchell, Courtney Houshmand, Golnaz Ghassemi, Marzyeh Killeen, Ronan Qiu, Wendy Hayden, Joel Rafiee, Farnaz Klochko, Chad Bevins, Nicholas Sazgara, Faeze Wong, S. Simon Moseley, Michael Halabi, Safwan Yeom, Kristen W. School of Medicine Stanford University United States Insight Centre for Data Analytics University College Dublin Ireland Icahn School of Medicine at Mount Sinai United States Koç University School of Medicine Turkey Division of Neuroradiology Amiralam Hospital Tehran University of Medical Sciences Iran Unity Health Toronto St. Michael’s Hospital University of Toronto Canada Amazon Web Services United States Universidade Federal de São Paulo Brazil Fleni Argentina Radiology Department Centre International Carthage Médicale Monastir Tunisia Hospital Israelita Albert Einstein SP São Paulo Brazil Department of Radiology Kyungpook National University Hospital Korea Republic of Department of Radiology St. Joseph’s Hospital and Medical Center United States Rajaie Cardiovascular Medical and Research Center Iran University of Medical Sciences Iran Dept. of Electrical Engineering and Computer Science at MIT United States Department of Radiology Henry Ford Health System DetroitMI United States Department of Radiology Guilan University of Medical Sciences Iran Department of Electrical Engineering Stanford University United States Ann & Robert H. Lurie Children's Hospital of Chicago United States

While it is well known that population differences from genetics, sex, race, and environmental factors contribute to disease, AI studies in medicine have largely focused on locoregional patient cohorts with less diverse data sources. Such limitation stems from barriers to large-scale data share and ethical concerns over data privacy. Federated learning (FL) is one potential pathway for AI development that enables learning across hospitals without data share. In this study, we show the results of various FL strategies on one of the largest and most diverse COVID-19 chest CT datasets: 21 participating hospitals across five continents that comprise >10,000 patients with >1 million images. We also propose an FL strategy that leverages synthetically generated data to overcome class and size imbalances. We also describe the sources of data heterogeneity in the context of FL, and show how even among the correctly labeled populations, disparities can arise due to these biases. Copyright © 2023, The Authors. All rights reserved.

关键词： Hospitals

Quantifying Human Activity Using Head Tracking

学校读者我要写书评

暂无评论

Quantifying Human Activity Using Head Tracking

IEEE International Conference on Bioinformatics and Biomedicine

作者： Alec M. Steele Melinda M. Bopp Zihang You Tanya S. Taylor Mehrdad Nourani Dennis H. Sullivan Dept. of Electrical & Computer Engineering University of Texas at Dallas Geriatric Research Edu. and Clinical Center Central Arkansas Veterans Healthcare System

ISBN: (纸本)9781728118680

This research focuses on quantifying human activity for healthcare applications. Distance traveled, average velocity, and postures are the primary focus. A Microsoft Kinect V2 camera was used to collect data for analysis. By tracking the human head, position relative to the camera can be determined. Accuracy of at least 87% was achieved for distance traveled and average velocity measurements. For posture detection, accuracy of at least 80% was achieved.

关键词： activity monitoring computer vision dept. mapping head tracking indoor location tracking localization realtime location tracking Human Activities images computer vision Average speed Cameras Quantify body position tracking

Exploration of Fine-Grained Parallelism for Load Balancing Eager K-truss on GPU and CPU

学校读者我要写书评

暂无评论

Exploration of Fine-Grained Parallelism for Load Balancing E...

IEEE Conference on High Performance Extreme Computing (HPEC)

作者： Mark Blanco Tze Meng Low Kyungjoo Kim Dept. of Electrical and Computer Engineering Carnegie Mellon University Pittsburgh United States Center for Computing Research Sandia National Laboratories Albuquerque United States

In this work we present a performance exploration on Eager K-truss, a linear-algebraic formulation of the K-truss graph algorithm. We address performance issues related to load imbalance of parallel tasks in symmetric, triangular graphs by presenting a fine-grained parallel approach to executing the support computation. This approach also increases available parallelism, making it amenable to GPU execution. We demonstrate our fine-grained parallel approach using implementations in Kokkos and evaluate them on an Intel Skylake CPU and an Nvidia Tesla V100 GPU. Overall, we observe between a 1.261. 48x improvement on the CPU and a 9.97-16.92x improvement on the GPU due to our fine-grained parallel formulation.

关键词： Task analysis Graphics processing units Parallel processing Load management Computational modeling Upper bound Sparse matrices

Roadmap on Deep Learning for Microscopy

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Volpe, Giovanni Wählby, Carolina Tian, Lei Hecht, Michael Yakimovich, Artur Monakhova, Kristina Waller, Laura Sbalzarini, Ivo F. Metzler, Christopher A. Xie, Mingyang Zhang, Kevin Lenton, Isaac C.D. Rubinsztein-Dunlop, Halina Brunner, Daniel Bai, Bijie Ozcan, Aydogan Midtvedt, Daniel Wang, Hao Sladoje, Nataša Lindblad, Joakim Smith, Jason T. Ochoa, Marien Barroso, Margarida Intes, Xavier Qiu, Tong Yu, Li-Yu You, Sixian Liu, Yongtao Ziatdinov, Maxim A. Kalinin, Sergei V. Sheridan, Arlo Manor, Uri Nehme, Elias Goldenberg, Ofri Shechtman, Yoav Moberg, Henrik K. Langhammer, Christoph Špačková, Barbora Helgadottir, Saga Midtvedt, Benjamin Argun, Aykut Thalheim, Tobias Cichos, Frank Bo, Stefano Hubatsch, Lars Pineda, Jesus Manzo, Carlo Bachimanchi, Harshith Selander, Erik Homs-Corbera, Antoni Fränzl, Martin de Haan, Kevin Rivenson, Yair Korczak, Zofia Adiels, Caroline Beck Mijalkov, Mite Veréb, Dániel Chang, Yu-Wei Pereira, Joana B. Matuszewski, Damian Kylberg, Gustaf Sintorn, Ida-Maria Caicedo, Juan C. Cimini, Beth A. Bell, Muyinatu Lediju A. Saraiva, Bruno M. Jacquemet, Guillaume Henriques, Ricardo Ouyang, Wei Le, Trang Gómez-De-Mariscal, Estibaliz Sage, Daniel Muñoz-Barrutia, Arrate Lindqvist, Ebba Josefson Bergman, Johanna Department of Physics University of Gothenburg Gothenburg41296 Sweden Dept. IT and SciLifeLab Uppsala University Sweden Department of Electrical and Computer Engineering Boston University BostonMA02215 United States Department of Biomedical Engineering Boston University BostonMA02215 United States Görlitz Germany Department of Renal Medicine Division of Medicine University College London Royal Free Hospital Campus London United Kingdom Artificial Intelligence for Life Sciences CIC Dorset United Kingdom Roche Pharma International Informatics Roche Diagnostics GmbH Mannheim Germany Department of Electrical Engineering and Computer Sciences UC Berkeley United States Technische Universität Dresden Faculty of Computer Science Dresden Germany Max Planck Institute of Molecular Cell Biology and Genetics Center for Systems Biology Dresden Dresden Germany University of Maryland College Park United States Klosterneuburg3400 Austria School of Mathematics and Physics The University of Queensland Brisbane4072 Australia Australian Research Council Centre of Excellence for Engineered Quantum Systems The University of Queensland Brisbane4072 Australia Département d’Optique P. M. Duffieux Institut FEMTO-ST Université Franche-Comté CNRS UMR 6174 Besançon France Department of Electrical and Computer Engineering University of California Los Angeles Los AngelesCA United States Department of Information Technology Uppsala University Sweden Department of Biomedical Engineering Rensselaer Polytechnic Institute TroyNY12180 United States Elephas 1 Erdman Pl. Madison53705 United States University of Wisconsin–Madison Department of Medical Physics MadisonWI53705 United States Department of Molecular and Cellular Physiology Albany Medical College AlbanyNY12208 United States Research Laboratory of Electronics Electrical Engineering and Computer Science Massachusetts Institute of Technology BostonMA United States Center for Nanophase Materials Sciences Oak R

Through digital imaging, microscopy has evolved from primarily being a means for visual observation of life at the micro- and nano-scale, to a quantitative tool with ever-increasing resolution and throughput. Artificial intelligence, deep neural networks, and machine learning are all niche terms describing computational methods that have gained a pivotal role in microscopy-based research over the past decade. This Roadmap is written collectively by prominent researchers and encompasses selected aspects of how machine learning is applied to microscopy image data, with the aim of gaining scientific knowledge by improved image quality, automated detection, segmentation, classification and tracking of objects, and efficient merging of information from multiple imaging modalities. We aim to give the reader an overview of the key developments and an understanding of possibilities and limitations of machine learning for microscopy. It will be of interest to a wide cross-disciplinary audience in the physical sciences and life sciences. Copyright © 2023, The Authors. All rights reserved.

关键词： Image enhancement