检索结果-内蒙古大学图书馆

Field-level simulation-based inference of galaxy clustering with convolutional neural networks

学校读者我要写书评

暂无评论

Physical Review D 2024年第8期109卷 083536-083536页

作者： Pablo Lemos Liam Parker ChangHoon Hahn Shirley Ho Michael Eickenberg Jiamin Hou Elena Massara Chirag Modi Azadeh Moradinezhad Dizgah Bruno Régaldo-Saint Blancard David Spergel Department of Physics Université de Montréal Montréal 1375 Avenue Thérèse-Lavoie-Roux Montréal QC H2V 0B3 Canada Mila—Quebec Artificial Intelligence Institute Montréal 6666 Rue Saint-Urbain Montréal QC H2S 3H1 Canada Ciela—Montreal Institute for Astrophysical Data Analysis and Machine Learning Montréal Canada Center for Computational Astrophysics Flatiron Institute 162 5th Avenue New York New York 10010 USA Department of Physics Princeton University Princeton New Jersey 08544 USA Center for Cosmology and Particle Physics Department of Physics New York University New York New York 10003 USA Department of Physics Carnegie Mellon University Pittsburgh Pennsylvania 15213 USA Center for Computational Mathematics Flatiron Institute 162 5th Avenue New York New York 10010 USA Department of Astronomy University of Florida 211 Bryant Space Science Center Gainesville Florida 32611 USA Max-Planck-Institut für Extraterrestrische Physik Postfach 1312 Giessenbachstrasse 1 85748 Garching bei München Germany Waterloo Centre for Astrophysics University of Waterloo 200 University Avenue W. Waterloo Ontario N2L 3G1 Canada Department of Physics and Astronomy University of Waterloo 200 University Avenue W. Waterloo Ontario N2L 3G1 Canada Département de Physique Théorique Université de Genève 24 quai Ernest Ansermet 1211 Genève 4 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 Pℓ, with analytic models based on perturbation theory. Consequently, they do not fully exploit the nonlinear and non-Gaussian features of the galaxy distribution. To address these limitations, we use the SimBIG forward modeling framework to perform SBI using normalizing flows. We apply SimBIG to a subset of the Baryon Oscillation Spectroscopic Survey 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 Ωm=0.267−0.029+0.033 and σ8=0.762−0.035+0.036. While our constraints on Ωm are in line with standard Pℓ analyses, ours 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 Pℓ. 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 the Dark Energy Spectroscopic Instrument, Prime Focus Spectrograph, and Euclid.

关键词： Cosmology Dark energy Dark matter Large scale structure of the Universe Astrophysical & cosmological simulations Bayesian methods Deep learning machine learning Numerical simulations in gravitation & astrophysics

Scaling provable adversarial defenses

学校读者我要写书评

暂无评论

arXiv 2018年

作者： Wong, Eric Metzen, Jan Hendrik Schmidt, Frank R. Zico Kolter, J. Machine Learning Department Carnegie Mellon University PittsburghPA15213 United States Bosch Center for Artificial Intelligence Renningen Germany Computer Science Department Carnegie Mellon University Bosch Center for Artificial Intelligence PittsburghPA15213 United States

Recent work has developed methods for learning deep network classifiers that are provably robust to norm-bounded adversarial perturbation;however, these methods are currently only possible for relatively small feedforward networks. In this paper, in an effort to scale these approaches to substantially larger models, we extend previous work in three main directions. First, we present a technique for extending these training procedures to much more general networks, with skip connections (such as ResNets) and general nonlinearities;the approach is fully modular, and can be implemented automatically (analogous to automatic differentiation). Second, in the specific case of `∞ adversarial perturbations and networks with ReLU nonlinearities, we adopt a nonlinear random projection for training, which scales linearly in the number of hidden units (previous approaches scaled quadratically). Third, we show how to further improve robust error through cascade models. On both MNIST and CIFAR data sets, we train classifiers that improve substantially on the state of the art in provable robust adversarial error bounds: from 5.8% to 3.1% on MNIST (with `∞ perturbations of ∊ = 0.1), and from 80% to 36.4% on CIFAR (with `∞ perturbations of ∊ = 2/255). Code for all experiments in the paper is available at https://***/locuslab/convex_adversarial/. Copyright © 2018, The Authors. All rights reserved.

关键词： Error analysis

A Novel Approach for Detection of Counterfeit Indian Currency Notes Using Deep Convolutional Neural Network

学校读者我要写书评

暂无评论

IOP Conference Series: Materials science and Engineering 2020年第2期981卷

作者： S Naresh Kumar Gaurav Singal Shwetha Sirikonda R. Nethravathi Department of Computer Science and Engineering SR Engineering College Warangal Telangana India Member Center for Artificial Intelligence and Deep Learning SR University Warangal Telangana India Department of Computer Science and Engineering Bennett University Greater Noida India Department of Computer Science and Engineering Sumathi Reddy Institute of Technology for Women Ananthasagar Hasanparthy Warangal India

In recent years, the Indian economy has shown rapid growth among all other major economies. India has been tragically reviled with issues like corruption and black currency, fake money notes is additionally major issues to it, in spite of a strong security feature are endorsed by RBI to print original currency. The advancement of color printing technology helped local racketeers and foreign racketeers to print a large amount of counterfeit Indian currency notes in the market. Albeit counterfeit money is being printed with accuracy, it likely is distinguished with some effort. In this paper, the proposed model efficiently detect the counterfeit Indian currency notes by adapting three-layered Deep Convolutional Neural Network (Deep ConvNet), and achieved an accuracy of 96.6%.

关键词：

Author Correction: APOE4 impairs myelination via cholesterol dysregulation in oligodendrocytes

学校读者我要写书评

暂无评论

Nature 2024年第8043期636卷 E9页

作者： Joel W Blanchard Leyla Anne Akay Jose Davila-Velderrain Djuna von Maydell Hansruedi Mathys Shawn M Davidson Audrey Effenberger Chih-Yu Chen Kristal Maner-Smith Ihab Hajjar Eric A Ortlund Michael Bula Emre Agbas Ayesha Ng Xueqiao Jiang Martin Kahn Cristina Blanco-Duque Nicolas Lavoie Liwang Liu Ricardo Reyes Yuan-Ta Lin Tak Ko Lea R'Bibo William T Ralvenius David A Bennett Hugh P Cam Manolis Kellis Li-Huei Tsai Picower Institute for Learning and Memory Massachusetts Institute of Technology Cambridge MA USA. Department of Brain and Cognitive Sciences Massachusetts Institute of Technology Cambridge MA USA. Nash Family Department of Neuroscience Black Family Stem Cell Institute Ronald M. Loeb Center for Alzheimer's Disease Friedman Brain Institute Icahn School of Medicine at Mt Sinai New York NY USA. MIT Computer Science and Artificial Intelligence Laboratory Cambridge MA USA. Human Technopole Milan Italy. Department of Neurobiology University of Pittsburgh Pittsburgh PA USA. Lewis-Sigler Institute for Integrative Genomics Princeton University Princeton NJ USA. Department of Medicine Emory Integrated Metabolomics and Lipidomics Core Emory University School of Medicine Atlanta GA USA. Department of Neurology Emory University School of Medicine Atlanta GA USA. Department of Biochemistry Emory University School of Medicine Atlanta GA USA. Rush Alzheimer's Disease Center Rush University Medical Center Chicago IL USA. MIT Computer Science and Artificial Intelligence Laboratory Cambridge MA USA. manoli@mit.edu. Broad Institute of Harvard and MIT Cambridge MA USA. manoli@mit.edu. Picower Institute for Learning and Memory Massachusetts Institute of Technology Cambridge MA USA. lhtsai@mit.edu. Department of Brain and Cognitive Sciences Massachusetts Institute of Technology Cambridge MA USA. lhtsai@mit.edu. Broad Institute of Harvard and MIT Cambridge MA USA. lhtsai@mit.edu.

The Bottleneck Simulator: A model-based deep reinforcement learning approach

学校读者我要写书评

暂无评论

arXiv 2018年

作者： Serban, Iulian Vlad Sankar, Chinnadhurai Pieper, Michael Pineau, Joelle Bengio, Yoshua Montreal Institute for Learning Algorithms Montreal Canada School of Computer Science McGill University Montreal and Facebook Artificial Intelligence Research

Deep reinforcement learning has recently shown many impressive successes. However, one major obstacle towards applying such methods to realworld problems is their lack of data-efficiency. To this end, we propose the Bottleneck Simulator: A model-based reinforcement learning method which combines a learned, factorized transition model of the environment with rollout simulations to learn an effective policy from few examples. The learned transition model employs an abstract, discrete (bottleneck) state, which increases sample efficiency by reducing the number of model parameters and by exploiting structural properties of the environment. We provide a mathematical analysis of the Bottleneck Simulator in terms of fixed points of the learned policy, which reveals how performance is affected by four distinct sources of error: An error related to the abstract space structure, an error related to the transition model estimation variance, an error related to the transition model estimation bias, and an error related to the transition model class bias. Finally, we evaluate the Bottleneck Simulator on two natural language processing tasks: A text adventure game and a real-world, complex dialogue response selection task. On both tasks, the Bottleneck Simulator yields excellent performance beating competing approaches. Copyright © 2018, The Authors. All rights reserved.

关键词： Deep learning

Topological Unwinding in an Exciton-Polariton Condensate Array

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Lyu, Guitao Minami, Yuki Kim, Na Young Byrnes, Tim Watanabe, Gentaro School of Physics Zhejiang Institute of Modern Physics Zhejiang University Zhejiang Hangzhou310027 China Division of Natural and Applied Sciences Duke Kunshan University Jiangsu Kunshan215300 China Faculty of Engineering Gifu University 1-1 Yanagido Gifu501-1193 Japan Institute for Quantum Computing University of Waterloo 200 University Ave. West WaterlooONN2L 3G1 Canada Department of Electrical and Computer Engineering University of Waterloo 200 University Ave. West WaterlooONN2L 3G1 Canada Waterloo Institute for Nanotechnology University of Waterloo 200 University Ave. West WaterlooONN2L 3G1 Canada New York University Shanghai NYU-ECNU Institute of Physics at NYU Shanghai Shanghai Frontiers Science Center of Artificial Intelligence and Deep Learning 567 West Yangsi Road Shanghai200126 China State Key Laboratory of Precision Spectroscopy School of Physical and Material Sciences East China Normal University Shanghai200062 China NYUAD Research Institute New York University Abu Dhabi United Arab Emirates Department of Physics New York University New YorkNY10003 United States Zhejiang Province Key Laboratory of Quantum Technology and Device Zhejiang University Zhejiang Hangzhou310027 China

The phase distribution in a Bose-Einstein condensate can realize various topological states classified by distinct winding numbers. While states with different winding numbers are topologically protected in the linear Schrödinger equation, when nonlinearities are introduced, violations of the topological protection can occur, leading to unwinding. Exciton-polariton condensates constitute a nonlinear open-dissipative system that is well suited to studying such physics. Here we show that a one-dimensional array of exciton-polariton condensates displays a spontaneous phase unwinding from a πto zero-state. We clarify that this collective mode transition is caused by the combined effect of nonlinearity and topological defects in the condensates. While the mode-switching phenomenon observed in our previous experiment was interpreted as the single-particle mode competition, we offer an alternative explanation in terms the collective phase unwinding and find its evidence by reanalyzing the experimental data. Our results open a route towards active control of the mode switching by manipulating the topological defects in prospective quantum polaritonic devices. Copyright © 2023, The Authors. All rights reserved.

关键词： Winding

Revisiting distillation and incremental classifier learning

学校读者我要写书评

暂无评论

arXiv 2018年

作者： Javed, Khurram Shafait, Faisal Deep Learning Laboratory National Center of Artificial Intelligence Islamabad Pakistan School of Electrical Engineering and Computer Science National University of Sciences and Technology

One of the key differences between the learning mechanism of humans and artificial Neural Networks (ANNs) is the ability of humans to learn one task at a time. ANNs, on the other hand, can only learn multiple tasks simultaneously. Any attempts at learning new tasks incrementally cause them to completely forget about previous tasks. This lack of ability to learn incrementally, called Catastrophic Forgetting, is considered a major hurdle in building a true AI system. In this paper, our goal is to isolate the truly effective existing ideas for incremental learning from those that only work under certain condi- tions. To this end, we first thoroughly analyze the current state of the art (iCaRL) method for incremental learning and demonstrate that the good performance of the system is not because of the reasons presented in the existing literature. We conclude that the success of iCaRL is primarily due to knowledge distillation and recognize a key limitation of knowledge distillation, i.e, it often leads to bias in classifiers. Finally, we propose a dynamic threshold moving algorithm that is able to success- fully remove this bias. We demonstrate the effectiveness of our algorithm on CIFAR100 and MNIST datasets showing near-optimal results. Our implementation is available at : https://***/Khurramjaved96/ incremental-learning. Copyright © 2018, The Authors. All rights reserved.

关键词： Neural networks

Evaluating the Accuracy and Reliability of Real-World Digital Mobility Outcomes in Older Adults After Hip Fracture: Cross-Sectional Observational Study

学校读者我要写书评

暂无评论

JMIR Formative Research 2025年 9卷 e67792页

作者： Berge, Martin A. Paraschiv-Ionescu, Anisoara Kirk, Cameron Küderle, Arne Micó-Amigo, Encarna Becker, Clemens Cereatti, Andrea Del Din, Silvia Engdal, Monika Garcia-Aymerich, Judith Grønvik, Karoline B. Hansen, Clint Hausdorff, Jeffrey M. Helbostad, Jorunn L. Jansen, Carl-Philipp Johnsen, Lars Gunnar Klenk, Jochen Koch, Sarah Maetzler, Walter Megaritis, Dimitrios Müller, Arne Rochester, Lynn Schwickert, Lars Taraldsen, Kristin Vereijken, Beatrix Department of Neuromedicine and Movement Science Norwegian University of Science and Technology Trondheim Norway Laboratory of Movement Analysis and Measurement Ecole Polytechnique Federale de Lausanne Lausanne Switzerland Translational and Clinical Research Institute Faculty of Medical Sciences Newcastle University Newcastle Upon Tyne United Kingdom Machine Learning and Data Analytics Lab Department Artificial Intelligence in Biomedical Engineering Friedrich-Alexander-Universität Erlangen-Nürnberg Erlangen Germany Geriatric Center Medical Faculty Heidelberg Heidelberg University Heidelberg Germany Department of Geriatrics and Rehabilitation Robert Bosch Hospital Stuttgart Germany Department of Electronics and Telecommunications Politecnico di Torino Turin Italy National Institute for Health and Care Research Newcastle Biomedical Research Centre Newcastle University The Newcastle Upon Tyne Hospitals NHS Foundation Trust Newcastle Upon Tyne United Kingdom Barcelona Institute for Global Health Barcelona Spain Department of Medicine and Life Sciences Universitat Pompeu Fabra Catalonia Barcelona Spain CIBER Epidemiología y Salud Pública Madrid Spain Department of Neurology University Hospital Schleswig-Holstein Kiel University Kiel Germany Center for the Study of Movement Cognition and Mobility Neurological Institute Tel Aviv Medical Center Tel Aviv Israel Department of Physical Therapy Faculty of Medical & Health Sciences Tel Aviv University Tel Aviv Israel Sagol School of Neuroscience Tel Aviv University Tel Aviv Israel Rush Alzheimer’s Disease Center Rush University Medical Center Chicago IL United States Department of Orthopedic Surgery Rush Medical College Rush University Chicago IL United States Department of Orthopaedic Surgery St. Olav’s Hospital Trondheim Norway Institute of Epidemiology and Medical Biometry Ulm University Ulm Germany IB University of Health and Social Sciences Study Centre Stuttgart Stuttgart Germany Department of Sport

Background: Algorithms estimating real-world digital mobility outcomes (DMOs) are increasingly validated in healthy adults and various disease cohorts. However, their accuracy and reliability in older adults after hip fracture, who often walk slowly for short durations, is underexplored. Objective: This study examined DMO accuracy and reliability in a hip fracture cohort considering walking bout (WB) duration, physical function, days since surgery, and walking aid use. Methods: In total, 19 community-dwelling participants were real-world monitored for 2.5 hours using a lower back wearable device and a reference system combining inertial modules, distance sensors, and pressure insoles. A total of 6 DMO estimates from 164 WBs from 58% (11/19) of the participants (aged 71-90 years;assessed 32-390 days after surgery;Short Physical Performance Battery [SPPB] scores of 3-12;gait speed range 0.39-1.34 m/s) were assessed against the reference system at the WB and participant level. We stratified by WB duration (all WBs, WBs of >10 seconds, WBs of 10-30 seconds, and WBs of >30 seconds) and lower versus higher SPPB scores and observed whether days since surgery and walking aid use affected DMO accuracy and reliability. Results: Across WBs, walking speed and distance ranged from 0.25 to 1.29 m/s and from 1.7 to 436.5 m, respectively. Estimation of walking speed, cadence, stride duration, number of steps, and distance stratified by WB duration showed intraclass correlation coefficients (ICCs) ranging from 0.50 to 0.99 and mean relative errors (MREs) from –6.9% to 12.8%. Stride length estimation showed poor reliability, with ICCs ranging from 0.30 to 0.49 and MREs from 6.1% to 13.2%. Walking speed and distance ICCs in the higher–SPPB score group ranged from 0.85 to 0.99, and MREs ranged from –10.1% to –1.7%. In the lower–SPPB score group, walking speed and distance ICCs ranged from 0.17 to 0.99, and MREs ranged from 13.5% to 32.6%. There was no discernible effect of time since s

关键词： active aging cadence digital mobility assessment distance free living gait walking speed wearable device

Predicting the clinical impact of human mutation with deep neural networks (vol 50, pg 1161, 2018)

学校读者我要写书评

暂无评论

NATURE GENETICS 2019年第2期51卷 364-364页

作者： Sundaram, Laksshman Gao, Hong Padigepati, Samskruthi Reddy McRae, Jeremy F. Li, Yanjun Kosmicki, Jack A. Fritzilas, Nondas Hakenberg, Jorg Dutta, Anindita Shon, John Xu, Jinbo Batzoglou, Serafim Li, Xiaolin Farh, Kyle Kai-How Illumina Artificial Intelligence Laboratory Illumina Inc San Diego USA Department of Computer Science Stanford University Stanford USA National Science Foundation Center for Big Learning University of Florida Gainesville USA Analytic and Translational Genetics Unit (ATGU) Department of Medicine Massachusetts General Hospital and Harvard Medical School Boston USA Toyota Technological Institute at Chicago Chicago USA

In the version of this article originally published, the name of author Serafim Batzoglou was misspelled. The error has been corrected in the HTML and PDF versions of the article.

关键词：