检索结果-内蒙古大学图书馆

Label-Efficient Self-Supervised Federated Learning for Tackling Data Heterogeneity in Medical Imaging

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Yan, Rui Qu, Liangqiong Wei, Qingyue Huang, Shih-Cheng Shen, Liyue Rubin, Daniel Xing, Lei Zhou, Yuyin The Institute for Computational and Mathematical Engineering Stanford University StanfordCA94305 United States The Department of Statistics and Actuarial Science The Institute of Data Science The University of Hong Kong 999077 Hong Kong The Department of Biomedical Data Science Stanford University StanfordCA94305 United States The Department of Electrical Engineering Stanford University StanfordCA94305 United States The Department of Biomedical Informatics Stanford University StanfordCA94305 United States The Department of Radiation Oncology Stanford University StanfordCA94305 United States The Department of Computer Science and Engineering University of California Santa CruzCA95064 United States

The collection and curation of large-scale medical datasets from multiple institutions is essential for training accurate deep learning models, but privacy concerns often hinder data sharing. Federated learning (FL) is a promising solution that enables privacy-preserving collaborative learning among different institutions, but it generally suffers from performance deterioration due to heterogeneous data distributions and a lack of quality labeled data. In this paper, we present a robust and label-efficient self-supervised FL framework for medical image analysis. Our method introduces a novel Transformer-based self-supervised pre-training paradigm that pre-trains models directly on decentralized target task datasets using masked image modeling, to facilitate more robust representation learning on heterogeneous data and effective knowledge transfer to downstream models. Extensive empirical results on simulated and real-world medical imaging non-IID federated datasets show that masked image modeling with Transformers significantly improves the robustness of models against various degrees of data heterogeneity. Notably, under severe data heterogeneity, our method, without relying on any additional pre-training data, achieves an improvement of 5.06%, 1.53% and 4.58% in test accuracy on retinal, dermatology and chest X-ray classification compared to the supervised baseline with ImageNet pre-training. In addition, we show that our federated self-supervised pre-training methods yield models that generalize better to out-of-distribution data and perform more effectively when fine-tuning with limited labeled data, compared to existing FL algorithms. The code is available at https://***/rui-yan/SSL-FL. © 2022, CC BY.

关键词： Deep learning

Reconstructing Suspended Sediment Concentrations in the Mekong River Basin Via Semi-Supervised-Based Deep Neural Networks

学校读者我要写书评

暂无评论

SSRN

SSRN 2024年

作者： Nguyen, Thi-Thu-Ha Vu, Duc Quang Doan, Ngoc Phu Thi Khanh Chi, Huynh Li, Peixin Van Binh, Doan An, Yimeng Tuan Dung, Pham Hoang, Tuan A. Thai Son, Mai Laboratory of Environmental Sciences and Climate Change Institute for Computational Science and Artificial Intelligence Van Lang University Ho Chi Minh City70000 Viet Nam Faculty of Environment School of Technology Van Lang University Ho Chi Minh City70000 Viet Nam Department of Computer Science and Information System Thai Nguyen University of Education Viet Nam School of Electronics Electrical Engineering and Computer Science Queen’s University Belfast United Kingdom College of Economics and Management Nanjing University of Aeronautics and Astronautics China Master program in Water Technology Reuse and Management Faculty of Engineering Vietnamese-German University Binh Duong Province Ben Cat Town820000 Viet Nam

The Mekong River Basin (MRB) is one of the highest biodiversity areas in the world and is the vital source of livelihoods for more than 60 millions of people over six countries in Southeast Asia. However, it is suffering from a severe sediment starvation problem caused by climate change and human activities, particularly dam constructions. Understanding long-term sediment changes thus plays a crucial role for management and contingency plans. Current sediment concentration data in the MRB are extremely sporadic, making it a non-trivial task to analyze them. This paper focuses on the sediment reconstruction to provide long-term reliable suspended sediment concentration (SSC) data using Machine Learning (ML) perspectives. In particular, a novel semi-supervised ML model, which can effectively deal with data sparsity, is introduced to predict/reconstruct missing sediment data. The key idea of this approach is to exploit abundant available hydroclimate data to reduce training overfitting rather than solely relying on sediment concentration data, thus enhancing the accuracy of the employed ML models. Extensive experiments on data obtained from 1979 to 2019 at the three main stations (i.e., Chiang Saen, Nong Khai, and Mukdahan) are conducted to demonstrate the performances of the proposed method compared to other state-of-the-art supervised techniques and traditional simulation methods commonly used in previous sediment forecasting/reconstruction studies. To the best of our knowledge, this approach is the first semi-supervised method to reconstruct sediment data in the field. Though this study focuses on the MRB, sediment sparsity is a very common problem in other river systems worldwide. Thus, this method can be widely applied for them for more effective sediment reconstruction/prediction in general. © 2024, The Authors. All rights reserved.

关键词： Biodiversity

Estimating Classification Confidence Using Kernel Densities

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Salamon, Peter Salamon, David Cantu, V. Adrian An, Michelle Perry, Tyler Edwards, Robert A. Segall, Anca M. The Department of Mathematics and Statistics The Computational Science Research Center The Viral Information Institute San Diego State University San DiegoCA92182 United States The Department of Mathematics and Statistics San Diego State University San DiegoCA92182 United States The Computational Science Research Center The Department of Biology San Diego State University San DiegoCA92182 United States The Perelman School of Medicine The University of Pennsylvania 425 Johnson Pavilion 3610 Hamilton Walk PhiladelphiaPA19104 United States The Computational Science Research Center San Diego State University San DiegoCA92182 United States Google 1600 Amphitheater Parkway Mountain ViewCA94043 United States The Bioinformatics and Medical Informatics Program San Diego State University San DiegoCA92182 United States Helix 9875 Towne Center Drive San DiegoCA92121 United States The Department of Biology The Computational Science Research Center The Viral Information Institute The Bioinformatics and Medical Informatics Program San Diego State University San DiegoCA92182 United States

This paper investigates the post-hoc calibration of confidence for "exploratory" machine learning classification problems. The difficulty in these problems stems from the continuing desire to push the boundaries of which categories have enough examples to generalize from when curating datasets, and confusion regarding the validity of those categories. We argue that for such problems the "one-versus-all" approach (top-label calibration) must be used rather than the "calibrate-the-full-response-matrix" approach advocated elsewhere in the literature. We introduce and test four new algorithms designed to handle the idiosyncrasies of category-specific confidence estimation. Chief among these methods is the use of kernel density ratios for confidence calibration including a novel, bulletproof algorithm for choosing the bandwidth. We test our claims and explore the limits of calibration on a bioinformatics application (PhANNs) [1], [2] as well as the classic MNIST benchmark [3]. Finally, our analysis argues that post-hoc calibration should always be performed, should be based only on the test dataset, and should be sanity-checked visually. Copyright © 2022, The Authors. All rights reserved.

关键词： Calibration

Current and future directions in network biology

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Zitnik, Marinka Li, Michelle M. Wells, Aydin Glass, Kimberly Gysi, Deisy Morselli Krishnan, Arjun Murali, T.M. Radivojac, Predrag Roy, Sushmita Baudot, Anaïs Bozdag, Serdar Chen, Danny Z. Cowen, Lenore Devkota, Kapil Gitter, Anthony Gosline, Sara Gu, Pengfei Guzzi, Pietro H. Huang, Heng Jiang, Meng Kesimoglu, Ziynet Nesibe Koyuturk, Mehmet Ma, Jian Pico, Alexander R. Pržulj, Nataša Przytycka, Teresa M. Raphael, Benjamin J. Ritz, Anna Sharan, Roded Shen, Yang Singh, Mona Slonim, Donna K. Tong, Hanghang Yang, Xinan Holly Yoon, Byung-Jun Yu, Haiyuan Milenković, Tijana Department of Biomedical Informatics Harvard University United States Department of Computer Science and Engineering University of Notre Dame United States Lucy Family Institute for Data and Society University of Notre Dame United States Eck Institute for Global Health University of Notre Dame United States Channing Division of Network Medicine Brigham and Women’s Hospital Harvard Medical School United States Department of Statistics Federal University of Paraná Brazil Department of Physics Northeastern University United States Department of Biomedical Informatics University of Colorado United States Department of Computer Science Virginia Tech United States Khoury College of Computer Sciences Northeastern University United States Department of Biostatistics and Medical Informatics University of Wisconsin-Madison United States Wisconsin Institute for Discovery United States Aix Marseille Université INSERM MMG Marseille France Department of Computer Science and Engineering University of North Texas United States Department of Mathematics University of North Texas United States Department of Computer Science Tufts University United States Morgridge Institute for Research United States Pacific Northwest National Laboratory United States Department of Medical and Surgical Sciences University Magna Graecia of Catanzaro Italy Department of Computer Science University of Maryland College Park United States National Center of Biotechnology Information National Library of Medicine National Institutes of Health Department of Computer and Data Sciences Case Western Reserve University United States Computational Biology Department School of Computer Science Carnegie Mellon University United States Institute of Data Science and Biotechnology Gladstone Institutes Department of Computer Science University College London United Kingdom ICREA Catalan Institution for Research and Advanced Studies Spain Department of Computer Science Princeton University United States Departmen

Network biology is an interdisciplinary field bridging computational and biological sciences that has proved pivotal in advancing the understanding of cellular functions and diseases across biological systems and scales. Although the field has been around for two decades, it remains nascent. It has witnessed rapid evolution, accompanied by emerging challenges. These challenges stem from various factors, notably the growing complexity and volume of data together with the increased diversity of data types describing different tiers of biological organization. We discuss prevailing research directions in network biology and highlight areas of inference and comparison of biological networks, multimodal data integration and heterogeneous networks, higher-order network analysis, machine learning on networks, and network-based personalized medicine. Following the overview of recent breakthroughs across these five areas, we offer a perspective on the future directions of network biology. Additionally, we offer insights into scientific communities, educational initiatives, and the importance of fostering diversity within the field. This paper establishes a roadmap for an immediate and long-term vision for network biology. © 2023, CC BY.

关键词： Machine learning

Benchmarking Noisy Intermediate Scale Quantum Error Mitigation Strategies for Ground State Preparation of the HCl Molecule

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Weaving, Tim Ralli, Alexis Kirby, William M. Love, Peter J. Succi, Sauro Coveney, Peter V. Centre for Computational Science Department of Chemistry University College London WC1H 0AJ United Kingdom Department of Physics and Astronomy Tufts University MedfordMA02155 United States Computational Science Initiative Brookhaven National Laboratory UptonNY11973 United States Center for Life Nano-Neuro Science @ La Sapienza Italian Institute of Technology Roma00161 Italy Department of Mechanical Engineering University College London WC1E 7JE United Kingdom Department of Physics Harvard University CambridgeMA02138 United States Advanced Research Computing Centre University College London WC1H 0AJ United Kingdom Informatics Institute University of Amsterdam Amsterdam1098 XH Netherlands

Due to numerous limitations including restrictive qubit topologies, short coherence times and prohibitively high noise floors, few quantum chemistry experiments performed on existing noisy intermediate-scale quantum hardware have achieved the high bar of chemical precision, namely energy errors to within 1.6 mHa of full configuration interaction. To have any hope of doing so, we must layer contemporary resource reduction techniques with best-in-class error mitigation methods;in particular, we combine the techniques of qubit tapering and the contextual subspace variational quantum eigensolver with several error mitigation strategies comprised of measurement-error mitigation, symmetry verification, zero-noise extrapolation and dual-state purification. We benchmark these strategies across a suite of eight 27-qubit IBM Falcon series quantum processors, taking preparation of the HCl molecule’s ground state as our testbed. Copyright © 2023, The Authors. All rights reserved.

关键词： Ground state

Contextual Subspace Variational Quantum Eigensolver Calculation of the Dissociation Curve of Molecular Nitrogen on a Superconducting Quantum Computer

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Weaving, Tim Ralli, Alexis Love, Peter J. Succi, Sauro Coveney, Peter V. Centre for Computational Science Department of Chemistry University College London WC1H 0AJ United Kingdom Department of Physics and Astronomy Tufts University MedfordMA02155 United States Computational Science Initiative Brookhaven National Laboratory UptonNY11973 United States Center for Life Nano-Neuro Science @ La Sapienza Italian Institute of Technology Roma00161 Italy Department of Mechanical Engineering University College London WC1E 7JE United Kingdom Department of Physics Harvard University CambridgeMA02138 United States Advanced Research Computing Centre University College London WC1H 0AJ United Kingdom Informatics Institute University of Amsterdam Amsterdam1098 XH Netherlands

In this work we present an experimental demonstration of the Contextual Subspace Variational Quantum Eigensolver on superconducting quantum hardware. In particular, we compute the potential energy curve for molecular nitrogen, where a dominance of static correlation in the dissociation limit proves challenging for many conventional quantum chemistry techniques. Our quantum simulations retain good agreement with the full configuration interaction energy in the chosen STO-3G basis, outperforming all benchmarked single-reference wavefunction techniques in capturing the bond-breaking appropriately. Moreover, our methodology is competitive with several multiconfigurational approaches, but at a considerable saving of quantum resource, meaning larger active spaces can be treated for a fixed qubit allowance. To achieve this result we deploy an error mitigation/suppression strategy comprised of dynamical decoupling, measurement-error mitigation and zero-noise extrapolation, in addition to circuit parallelization that not only provides passive averaging of noise but improves the effective shot-yield to reduce the measurement overhead. Furthermore, we introduce a modification to previous adaptive ansatz construction algorithms that incorporates hardware-awareness into our variational circuits to minimize the transpilation cost for the target qubit topology. Copyright © 2023, The Authors. All rights reserved.

关键词： Qubits

Development of Educational Game for Indonesian History Lesson Focusing on Indonesia Pre-Literacy Relics 'Prehistoria' for Grade X High School Student

学校读者我要写书评

暂无评论

Development of Educational Game for Indonesian History Lesso...

2020 International Conference on Computer engineering, Network, and Intelligent Multimedia, CENIM 2020

作者： Nugroho, Supeno Mardi Susiki Rahman, Muhammad Daffa Abiyyu Hariadi, Mochamad Arief, Muhammad Faculty of Intelligent Electrical and Informatics Technology Sepuluh Nopember Institute of Technology Computer Engineering Department Surabaya Indonesia Tokyo University of Technology Media Science Program Graduate School of Bionics Computer and Media Sciences Tokyo Japan

ISBN: (纸本)9781728182834

The development of game technology has opened up new possibilities in learning methods through educational games. However, in Indonesia, the application of this concept is lacking. One lesson that can benefit from educational games is Indonesian History in class X, with a focus on history in the pre-literacy period. This study aims to create a puzzle-Adventure type of digital game that can help educate students about the pre-literacy period, developed with Unity, and uses WebGL as a target device while referring to Indonesia's revised 2013 national curriculum. The result of this research is the creation of a digital game that can be played on average of 254.36 seconds, with potential for educational use. A survey result shows that many of the ancient objects and some of the ages are well-recognized. However, the survey also shows many of the gameplay aspects, such as UI, gameplay features, and adventure elements, need improvements. © 2020 IEEE.

关键词： Surveys

Digital Nomads during the COVID-19 Pandemic: Evidence from Narratives on Reddit discussions

学校读者我要写书评

暂无评论

Digital Nomads during the COVID-19 Pandemic: Evidence from N...

International Conference on Computer Supported Cooperative Work in Design

作者： Marcos Antonio De Almeida Antó nio Correia Jano Moreira De Souza Daniel Schneider Systems Engineering and Computer Science Program (PESC/COPPE/UFRJ) Rio de Janeiro Brazil INESC TEC and University of Tr&#x00E1 s-os-Montes e Alto Douro Vila Real Portugal Postgraduate Program in Informatics PPGI/UFRJ Rio de Janeiro Brazil NCE/UFRJ T&#x00E9 rcio Pacitti Institute of Computer Applications and Research Rio de Janeiro Brazil

In this paper, we report on new findings about the results of an empirical study which aims to investigate how the COVID-19 pandemic has been shaping nomadic work practices and also challenging the lifestyles of digital nomads (DN). To do this, we collected textual data from posts in a Reddit community. We argue that, in order to understand how to design technical solutions for the so-called ‘new normal’ working conditions, one way to approach this is to understand how digital nomads are being impacted in their work practices and routines, and also how they are seeing the future of their technology-mediated work-life space. Finally, we show how evidence collected from DNs about their experiences and difficulties perceived during the pandemic period can inform CSCW researchers worldwide about future design-oriented strands.

关键词：

Gga + U Studies on Doping-Induced Modulation of the Electronic and Optical Properties in Na2znp2o7 Compound

学校读者我要写书评

暂无评论

SSRN

SSRN 2023年

作者： Hailouf, Houssam Eddine Obodo, Kingsley O. Gueddim, Ahmed Gacem, Lakhdar Bouhafs, Bachir Reggab, Khalid Materials Science and Informatics Laboratory Faculty of Science University of Djelfa Djelfa17000 Algeria HySA Infrastructure Centre of Competence Faculty of Engineering North-West University Private Bag X6001 Potchefstroom2531 South Africa National Institute of Theoretical and Computational Sciences Johannesburg2000 South Africa Laboratoire de Modélisation et Simulation en Sciences des Matériaux Université Djillali Liabès de Sidi Bel-Abbès Sidi Bel-Abbès22000 Algeria

In this study, we present a comprehensive analysis of the optical properties of a Na2ZnP2O7 host matrix doped with Ti and Sn ions, employing ab-initio density functional theory (DFT) calculations. Our investigation explores the impact of dopant ions on the properties of the host matrix. We utilize the GGA and GGA+U functional to determine the structural, energetic, electronic, and optical characteristics. The results demonstrate the energetic favorability of Ti and Sn doping, leading to the formation of a unique crystal structure that holds promise for diverse applications. Notably, intriguing electronic and optical features are observed within the electronic band gap of host due to the presence of Ti and Sn dopants. These features result in modifications of the electronic band structure, and dielectric functions. Moreover, our investigation reveals the nonmagnetic nature of Sn-doped host, as indicated by a zero total magnetic moment. In contrast, Ti-doped host exhibits a ferromagnetic character. © 2023, The Authors. All rights reserved.

关键词： Crystal structure