检索结果-内蒙古大学图书馆

SimBIG: Field-level Simulation-Based Inference of Galaxy Clustering

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Lemos, Pablo Parker, Liam Hahn, ChangHoon Ho, Shirley Eickenberg, Michael Hou, Jiamin Massara, Elena Modi, Chirag Dizgah, Azadeh Moradinezhad Blancard, Bruno Régaldo-Saint Spergel, David Department of Physics Université de Montréal 1375 Avenue Thérèse-Lavoie-Roux MontréalQCH2V 0B3 Canada Mila - Quebec Artificial Intelligence Institute 6666 Rue Saint-Urbain MontréalQCH2S 3H1 Canada Ciela - Montreal Institute for Astrophysical Data Analysis and Machine Learning Montréal Canada Center for Computational Astrophysics Flatiron Institute 162 5th Avenue New YorkNY10010 United States Department of Physics Princeton University PrincetonNJ08544 United States Center for Cosmology and Particle Physics Department of Physics New York University New YorkNY10003 United States Department of Physics Carnegie Mellon University PittsburghPA15213 United States Center for Computational Mathematics Flatiron Institute 162 5th Avenue New YorkNY10010 United States Department of Astronomy University of Florida 211 Bryant Space Science Center GainesvilleFL32611 United States Max-Planck-Institut für Extraterrestrische Physik Postfach 1312 Giessenbachstrasse 1 Garching bei München85748 Germany Waterloo Centre for Astrophysics University of Waterloo 200 University Ave W WaterlooONN2L 3G1 Canada Department of Physics and Astronomy University of Waterloo 200 University Ave W WaterlooONN2L 3G1 Canada Département de Physique Théorique Université de Genève 24 quai Ernest Ansermet Genève 41211 Switzerland

We present the first simulation-based inference (SBI) of cosmological parameters from field-level analysis of galaxy clustering. Standard galaxy clustering analyses rely on analyzing summary statistics, such as the power spectrum, P, with analytic models based on perturbation theory. Consequently, they do not fully exploit the non-linear and non-Gaussian features of the galaxy distribution. To address these limitations, we use the SimBIG forward modelling framework to perform SBI using normalizing flows. We apply SimBIG to a subset of the BOSS CMASS galaxy sample using a convolutional neural network with stochastic weight averaging to perform massive data compression of the galaxy field. We infer constraints on (Equation presented) and (Equation presented). While our constraints on Ωm are in-line with standard P analyses, those on σ8 are 2.65× tighter. Our analysis also provides constraints on the Hubble constant H0 = 64.5 ± 3.8 km/s/Mpc from galaxy clustering alone. This higher constraining power comes from additional non-Gaussian cosmological information, inaccessible with P. We demonstrate the robustness of our analysis by showcasing our ability to infer unbiased cosmological constraints from a series of test simulations that are constructed using different forward models than the one used in our training dataset. This work not only presents competitive cosmological constraints but also introduces novel methods for leveraging additional cosmological information in upcoming galaxy surveys like DESI, PFS, and Euclid. © 2023, CC0.

关键词： Galaxies

Time series featurization via topological data analysis

学校读者我要写书评

暂无评论

arXiv 2018年

作者： Kim, Kwangho Kim, Jisu Rinaldo, Alessandro Department of Statistics & Data Science Machine Learning Department} Carnegie Mellon University PittsburghPA15213 United States Inria Saclay – Île-de-France Palaiseau France Department of Statistics & Data Science Carnegie Mellon University Pittsburgh United States

We develop a novel algorithm for feature extraction in time series data by leveraging tools from topological data analysis. Our algorithm provides a simple, efficient way to successfully harness topological features of the attractor of the underlying dynamical system for an observed time series. The proposed methodology relies on the persistent landscapes and silhouette of the Rips complex obtained after a de-noising step based on principal components applied to a time-delayed embedding of a noisy, discrete time series sample. We analyze the stability properties of the proposed approach and show that the resulting TDA-based features are robust to sampling noise. Experiments on synthetic and real-world data demonstrate the effectiveness of our approach. We expect our method to provide new insights on feature extraction from granular, noisy time series data. Copyright © 2018, The Authors. All rights reserved.

关键词： Time series

Hybrid physical-deep learning model for astronomical inverse problems

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Lanusse, François Melchior, Peter Moolekamp, Fred Berkeley Center for Cosmological Physics Berkeley Institute for Data Science University of California Berkeley BerkeleyCA94709 United States Department of Astrophysical Sciences Center for Statistics and Machine Learning Princeton University PrincetonNJ08544 United States LSST Project Management Office TucsonAZ United States Department of Astrophysical Sciences Princeton University PrincetonNJ08544 United States

We present a Bayesian machine learning architecture that combines a physically motivated parametrization and an analytic error model for the likelihood with a deep generative model providing a powerful data-driven prior for complex signals. This combination yields an interpretable and differentiable generative model, allows the incorporation of prior knowledge, and can be utilized for observations with different data quality without having to retrain the deep network. We demonstrate our approach with an example of astronomical source separation in current imaging data, yielding a physical and interpretable model of astronomical scenes. Copyright © 2019, The Authors. All rights reserved.

关键词： Source separation

Algorithmic Regularization in learning Deep Homogeneous Models: Layers are Automatically Balanced

学校读者我要写书评

暂无评论

arXiv 2018年

作者： Du, Simon S. Hu, Wei Lee, Jason D. Machine Learning Department School of Computer Science Carnegie Mellon University Computer Science Department Princeton University. Department of Data Sciences and Operations Marshall School of Business University of Southern California.

We study the implicit regularization imposed by gradient descent for learning multi-layer homogeneous functions including feed-forward fully connected and convolutional deep neural networks with linear, ReLU or Leaky ReLU activation. We rigorously prove that gradient flow (i.e. gradient descent with infinitesimal step size) effectively enforces the differences between squared norms across different layers to remain invariant without any explicit regularization. This result implies that if the weights are initially small, gradient flow automatically balances the magnitudes of all layers. Using a discretization argument, we analyze gradient descent with positive step size for the non-convex low-rank asymmetric matrix factorization problem without any regularization. Inspired by our findings for gradient flow, we prove that gradient descent with step sizes ηt = O (t -(1/2 +δ)) (0 Copyright © 2018, The Authors. All rights reserved.

关键词： Gradient methods

Distribution-Free Prediction Sets for Two-Layer Hierarchical Models

学校读者我要写书评

暂无评论

arXiv 2018年

作者： Dunn, Robin Wasserman, Larry Ramdas, Aaditya Novartis Pharmaceuticals Corporation Advanced Methodology and Data Science East HanoverNJ United States Department of Statistics & Data Science Carnegie Mellon University PittsburghPA United States Machine Learning Department Carnegie Mellon University PittsburghPA United States

We consider the problem of constructing distribution-free prediction sets for data from two-layer hierarchical distributions. For iid data, prediction sets can be constructed using the method of conformal prediction. The validity of conformal prediction hinges on the exchangeability of the data, which does not hold when groups of observations come from distinct distributions, such as multiple observations on each patient in a medical database. We extend conformal methods to a hierarchical setting. We develop CDF pooling, single subsampling, and repeated subsampling approaches to construct prediction sets in unsupervised and supervised settings. We compare these approaches in terms of coverage and average set size. If asymptotic coverage is acceptable, we recommend CDF pooling for its balance between empirical coverage and average set size. If we desire coverage guarantees, then we recommend the repeated subsampling approach. Supplementary materials are available online. Copyright © 2018, The Authors. All rights reserved.

关键词： Forecasting

FUTURE-AI: International consensus guideline for trustworthy and deployable artificial intelligence in healthcare

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Lekadir, Karim Feragen, Aasa Fofanah, Abdul Joseph Frangi, Alejandro F. Buyx, Alena Emelie, Anais Lara, Andrea Porras, Antonio R. Chan, An-Wen Navarro, Arcadi Glocker, Ben Botwe, Benard O. Khanal, Bishesh Beger, Brigit Wu, Carol C. Cintas, Celia Langlotz, Curtis P. Rueckert, Daniel Mzurikwao, Deogratias Fotiadis, Dimitrios I. Zhussupov, Doszhan Ferrante, Enzo Meijering, Erik Weicken, Eva González, Fabio A. Asselbergs, Folkert W. Prior, Fred Krestin, Gabriel P. Collins, Gary S. Tegenaw, Geletaw S. Kaissis, Georgios Misuraca, Gianluca Tsakou, Gianna Dwivedi, Girish Kondylakis, Haridimos Jayakody, Harsha Woodruf, Henry C. Mayer, Horst Joachim Aerts, Hugo JWL Walsh, Ian Chouvarda, Ioanna Buvat, Irène Tributsch, Isabell Rekik, Islem Duncan, James Kalpathy-Cramer, Jayashree Zahir, Jihad Park, Jinah Mongan, John Gichoya, Judy W. Schnabel, Julia A. Kushibar, Kaisar Riklund, Katrine Mori, Kensaku Marias, Kostas Amugongo, Lameck M. Fromont, Lauren A. Maier-Hein, Lena Alberich, Leonor Cerdá Rittner, Leticia Phiri, Lighton Marrakchi-Kacem, Linda Donoso-Bach, Lluís Martí-Bonmatí, Luis Cardoso, M. Jorge Bobowicz, Maciej Shabani, Mahsa Tsiknakis, Manolis Zuluaga, Maria A. Bielikova, Maria Fritzsche, Marie-Christine Camacho, Marina Linguraru, Marius George Wenzel, Markus De Bruijne, Marleen Tolsgaard, Martin G. Ghassemi, Marzyeh Ashrafuzzaman, Md Goisauf, Melanie Yaqub, Mohammad Abadía, Mónica Cano Mahmoud, Mukhtar M.E. Elattar, Mustafa Rieke, Nicola Papanikolaou, Nikolaos Lazrak, Noussair Díaz, Oliver Salvado, Olivier Pujol, Oriol Sall, Ousmane Guevara, Pamela Gordebeke, Peter Lambin, Philippe Brown, Pieta Abolmaesumi, Purang Dou, Qi Lu, Qinghua Osuala, Richard Nakasi, Rose Zhou, S. Kevin Napel, Sandy Colantonio, Sara Albarqouni, Shadi Joshi, Smriti Carter, Stacy Klein, Stefan Petersen, Steffen E. Aussó, Susanna Awate, Suyash Raviv, Tammy Riklin Cook, Tessa Mutsvangwa, Tinashe E.M. Rogers, Wendy A. Niessen, Wiro J. Puig-Bosch, Xènia Zeng, Yi Mohammed, Yunusa G. Aquino, Yves Saint James Salahuddin, Zohaib Starmans, Martijn P.A. Department de Matemàtiques i Informàtica Universitat de Barcelona Barcelona Spain Barcelona Spain DTU Compute Technical University of Denmark Kgs Lyngby Denmark Department of Mathematics and Computer Science Faculty of Science and Technology Milton Margai Technical University Freetown Sierra Leone Center for Computational Imaging & Simulation Technologies in Biomedicine Schools of Computing and Medicine University of Leeds Leeds United Kingdom Cardiovascular Science and Electronic Engineering Departments KU Leuven Leuven Belgium Institute of History and Ethics in Medicine Technical University of Munich Munich Germany Faculty of Engineering of Systems Informatics and Sciences of Computing Galileo University Guatemala City Guatemala Department of Biostatistics and Informatics Colorado School of Public Health University of Colorado Anschutz Medical Campus AuroraCO United States Department of Medicine Women’s College Research Institute University of Toronto Toronto Canada Universitat Pompeu Fabra BarcelonaBeta Brain Research Center Barcelona Spain Department of Computing Imperial College London London United Kingdom School of Biomedical & Allied Health Sciences University of Ghana Accra Ghana Department of Midwifery & Radiography School of Health & Psychological Sciences City University of London United Kingdom Kathmandu Nepal European Heart Network Brussels Belgium Department of Thoracic Imaging University of Texas MD Anderson Cancer Center Houston United States IBM Research Africa Nairobi Kenya Departments of Radiology Medicine and Biomedical Data Science Stanford University School of Medicine Stanford United States Institute for AI and Informatics in Medicine Klinikum rechts der Isar Technical University Munich Munich Germany Department of Computing Imperial College London London United Kingdom Muhimbili University of Health and Allied Sciences Dar es Salaam Tanzania United Republic of Ioannina Greece Almaty AI Lab Almaty Kazakhstan

Background: Despite major advances in artificial intelligence (AI) research for healthcare, the deployment and adoption of AI technologies remain limited in clinical practice. In recent years, concerns have been raised about the clinical, technical, ethical and legal risks associated with healthcare AI. To increase adoption in the real world, it is essential that AI technologies are trusted and accepted by patients, clinicians, health organisations and authorities. This paper describes the FUTURE-AI framework as the first international consensus guideline for trustworthy AI in healthcare. Methods: The FUTURE-AI consortium was founded in 2021 and now comprises 117 interdisciplinary experts from 50 countries representing all continents, including AI scientists, clinical researchers, biomedical ethicists, and social scientists. Over a two-year period, the consortium established guiding principles and best practices for trustworthy and deployable AI through an iterative process comprising an in-depth literature review, a modified Delphi survey, and online consensus meetings. Findings: The FUTURE-AI framework was established based on six guiding principles for trustworthy AI in healthcare, i.e. Fairness, Universality, Traceability, Usability, Robustness and Explainability. Through consensus, a set of 30 best practices were defined, addressing technical, clinical, socio-ethical and legal dimensions of trustworthy AI. The recommendations cover the entire lifecycle of healthcare AI, from design, development and validation to regulation, deployment, and monitoring. Interpretation: FUTURE-AI is a structured, risk-informed framework which provides guidance for constructing healthcare AI tools that will be trusted, deployed and adopted in real-world clinical practice. Researchers are encouraged to take the recommendations into account in proof-of-concept stages to facilitate future translation towards clinical practice of healthcare AI. Copyright © 2023, The Authors. All rights

关键词： Clinical research

On ADMM in deep learning: Convergence and saturation-avoidance

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Zeng, Jinshan Lin, Shao-Bo Yao, Yuan Zhou, Ding-Xuan School of Computer and Information Engineering Jiangxi Normal University Nanchang China Liu Bie Ju Centre for Mathematical Sciences City University of Hong Kong Hong Kong Hong Kong Department of Mathematics Hong Kong University of Science and Technology Hong Kong Hong Kong Center of Intelligent Decision-Making and Machine Learning School of Management Xi’an Jiaotong University Xi’an China School of Data Science Department of Mathematics City University of Hong Kong Hong Kong Hong Kong

In this paper, we develop an alternating direction method of multipliers (ADMM) for deep neural networks training with sigmoid-type activation functions (called sigmoid-ADMM pair), mainly motivated by the gradient-free nature of ADMM in avoiding the saturation of sigmoid-type activations and the advantages of deep neural networks with sigmoid-type activations (called deep sigmoid nets) over their rectified linear unit (ReLU) counterparts (called deep ReLU nets) in terms of approximation. In particular, we prove that the approximation capability of deep sigmoid nets is not worse than that of deep ReLU nets by showing that ReLU activation function can be well approximated by deep sigmoid nets with two hidden layers and finitely many free parameters but not vice-verse. We also establish the global convergence of the proposed ADMM for the nonlinearly constrained formulation of the deep sigmoid nets training from arbitrary initial points to a Karush-Kuhn-Tucker (KKT) point at a rate of order O(1/k). Besides sigmoid activation, such a convergence theorem holds for a general class of smooth activations. Compared with the widely used stochastic gradient descent (SGD) algorithm for the deep ReLU nets training (called ReLU-SGD pair), the proposed sigmoid-ADMM pair is practically stable with respect to the algorithmic hyperparameters including the learning rate, initial schemes and the pro-processing of the input data. Moreover, we find that to approximate and learn simple but important functions the proposed sigmoid-ADMM pair numerically outperforms the ReLU-SGD pair. Copyright © 2019, The Authors. All rights reserved.

关键词： Chemical activation

Sequential monte carlo method for bayesian multiple testing of pairwise interactions among large number of neurons 14

学校读者我要写书评

暂无评论

Sequential monte carlo method for bayesian multiple testing ...

14th International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery, ICNC-FSKD 2018

作者： Liu, Bin Vinci, Giuseppe Snyder, Adam C. Kass, Robert E. School of Computer Science Jiangsu Key Lab of Big Data Security Intelligent Processing Nanjing University of Posts and Telecommunications Nanjing210023 China Department of Statistics Rice University Houston United States Visual Neuroscience Lab University of Pittsburgh Pittsburgh United States Basis of Cognition Carnegie Mellon University Machine Learning Department Center for the Neural Department of Statistics Pittsburgh United States

ISBN: (纸本)9781538680971

The problem of multiple testing arises in many contexts, including testing for pairwise interaction among a large number of neurons. Recently a method was developed to control false positives when covariate information, such as distances between pairs of neurons, is available. This method, however, relies on computationally-intensive Markov Chain Monte Carlo (MCMC). Here we develop an alternative, based on Sequential Monte Carlo, which only requires one pass of the data. This scheme considers data items sequentially, with relevant probabilities being updated at each step. Simulation experiments demonstrate that the proposed algorithm delivers results as accurately as the previous MCMC method. We illustrate the method by using it to analyze neural recordings from extrastriate cortex in a macaque monkey. © 2018 IEEE.

关键词： Neurons

Modern applications of machine learning in quantum sciences

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Dawid, Anna Arnold, Julian Requena, Borja Gresch, Alexander Plodzien, Marcin Donatella, Kaelan Nicoli, Kim A. Stornati, Paolo Koch, Rouven Büttner, Miriam Okula, Robert Muñoz–Gil, Gorka Vargas–Hernández, Rodrigo A. Cervera-Lierta, Alba Carrasquilla, Juan Dunjko, Vedran Gabrié, Marylou Huembeli, Patrick van Nieuwenburg, Evert Vicentini, Filippo Wang, Lei Wetzel, Sebastian J. Carleo, Giuseppe Greplová, Eliška Krems, Roman Marquardt, Florian Tomza, Michal Lewenstein, Maciej Dauphin, Alexandre Faculty of Physics University of Warsaw Poland ICFO - Institut de Ciències Fotòniques The Barcelona Institute of Science and Technology Castelldefels Barcelona08860 Spain Center for Computational Quantum Physics Flatiron Institute New York United States Department of Physics University of Basel Switzerland Institute for Theoretical Physics Heinrich Heine University Düsseldorf Germany Institute for Quantum Inspired and Quantum Optimization Hamburg University of Technology Germany Université de Paris CNRS Laboratoire Matériaux et Phénomènes Quantiques France Machine Learning Group Technische Universität Berlin Germany BIFOLD Berlin Institute for the Foundations of Learning and Data Berlin10587 Germany Department of Applied Physics Aalto University Espoo Finland Institute of Physics Albert-Ludwig University of Freiburg Germany International Centre for Theory of Quantum Technologies University of Gdańsk Poland Department of Algorithms and System Modeling Faculty of Electronics Faculty of Electronics Telecommunications and Informatics Gdańsk University of Technology Poland Institute for Theoretical Physics University of Innsbruck Austria Department of Chemistry University of Toronto Canada Vector Institute for Artificial Intelligence MaRS Centre Toronto Canada Department of Chemistry and Chemical Biology McMaster University Hamilton Canada Barcelona Supercomputing Center Spain LIACS Leiden University Netherlands CMAP École Polytechnique France Switzerland Menten AI Inc. Palo AltoCA United States Niels Bohr Institute Copenhagen Denmark CPHT CNRS École Polytechnique Institut Polytechnique de Paris PalaiseauF-91128 France Beijing National Lab for Condensed Matter Physics Institute of Physics Chinese Academy of Sciences Beijing China Songshan Lake Materials Laboratory Dongguan China Perimeter Institute for Theoretical Physics Waterloo Canada Kavli Institute of Nanoscience Delft University of Technology DelftNL-2600 GA Netherlands Department of

In this book, we provide a comprehensive introduction to the most recentadvances in the application of machine learning methods in quantum sciences. Wecover the use of deep learning and kernel methods in supervised, unsupervised,and reinforcement learning algorithms for phase classification, representationof many-body quantum states, quantum feedback control, and quantum circuitsoptimization. Moreover, we introduce and discuss more specialized topics suchas differentiable programming, generative models, statistical approach tomachine learning, and quantum machine learning. © 2022, CC BY.

关键词： Deep learning