检索结果-内蒙古大学图书馆

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 2018年 10742 LNCS卷

作者： Voronkov, Andrei Petrenko, Alexander K. The University of Manchester Manchester United Kingdom Ivannikov Institute for System Programming of RAS Moscow Russia

来源：评论

学校读者我要写书评

暂无评论

Scientific Data Lake for High Luminosity LHC project and other data-intensive particle and astro-particle physics experiments 5

Scientific Data Lake for High Luminosity LHC project and oth...

引用

5th International Conference on Particle Physics and Astrophysics, ICPPA 2020

作者： Alekseev, A. Kiryanov, A. Klimentov, A. Korchuganova, T. Mitsyn, V. Oleynik, D. Smirnov, A. Smirnov, S. Zarochentsev, A. Compiler Technology Department Ivannikov Institute for System Programming of the Ras Moscow109004 Russia Cloud Technologies and Big Data Analytics Laboratory Plekhanov Russian University of Economics Moscow117997 Russia Atlas Group Universidad Andres Bello Santiago7550196 Chile Department of Computing Resources and Technologies Nrc "kurchatov Institute"-PNPI Gatchina188300 Russia Physics Department Brookhaven National Laboratory UptonNY11973 United States Laboratory of Information Technologies Joint Institute for Nuclear Research Dubna141980 Russia Institute of Nuclear Physics and Engineering National Research Nuclear University MEPhI Moscow115409 Russia Laboratory of Ultra-High Energy Physics Saint Petersburg State University Saint Petersburg198504 Russia

The next phase of LHC Operations-High Luminosity LHC (HL-LHC), which is aimed at ten-fold increase in the luminosity of proton-proton collisions at the energy of 14 TeV, is expected to start operation in 2027-2028 and will deliver an unprecedented scientific data volume of multi-exabyte scale. This amount of data has to be stored and the corresponding storage system should ensure fast and reliable data delivery for processing by scientific groups distributed all over the world. The present LHC computing and data processing model will not be able to provide the required infrastructure growth even taking into account the expected hardware technology evolution. To address this challenge the new state-of-The-Art computing infrastructure technologies are now being developed and are presented here. The possibilities of application of the HL-LHC distributed data handling technique for other particle and astro-particle physics experiments dealing with large-scale data volumes like DUNE, LSST, Belle-II, JUNO, SKAO etc. are also discussed. © Published under licence by IOP Publishing Ltd.

关键词： Luminance

来源：评论

学校读者我要写书评

暂无评论

High-probability convergence for composite and distributed stochastic minimization and variational inequalities with heavy-tailed noise 24

High-probability convergence for composite and distributed s...

引用

Proceedings of the 41st International Conference on Machine Learning

作者： Eduard Gorbunov Abdurakhmon Sadiev Marina Danilova Samuel Horváth Gauthier Gidel Pavel Dvurechensky Alexander Gasnikov Peter Richtárik Mohamed bin Zayed University of Artificial Intelligence UAE King Abdullah University of Science and Technology KSA Moscow Institute of Physics and Technology Russia Université de Montréal and Mila Canada and Canada CIFAR AI Chair Weierstrass Institute for Applied Analysis and Stochastics Germany University Innopolis Russia and Ivannikov Institute for System Programming RAS Russia and Moscow Institute of Physics and Technology Russia and Skolkovo Institute of Science and Technology Russia

High-probability analysis of stochastic first-order optimization methods under mild assumptions on the noise has been gaining a lot of attention in recent years. Typically, gradient clipping is one of the key algorithmic ingredients to derive good high-probability guarantees when the noise is heavy-tailed. However, if implemented naïvely, clipping can spoil the convergence of the popular methods for composite and distributed optimization (Prox-SGD/Parallel SGD) even in the absence of any noise. Due to this reason, many works on high-probability analysis consider only unconstrained non-distributed problems, and the existing results for composite/distributed problems do not include some important special cases (like strongly convex problems) and are not optimal. To address this issue, we propose new stochastic methods for composite and distributed optimization based on the clipping of stochastic gradient differences and prove tight high-probability convergence results (including nearly optimal ones) for the new methods. In addition, we also develop new methods for composite and distributed variational inequalities and analyze the high-probability convergence of these methods.

关键词：

来源：评论

学校读者我要写书评

暂无评论

High-Probability Convergence for Composite and Distributed Stochastic Minimization and Variational Inequalities with Heavy-Tailed Noise

arXiv

引用

arXiv 2023年

作者： Gorbunov, Eduard Sadiev, Abdurakhmon Danilova, Marina Horváth, Samuel Gidel, Gauthier Dvurechensky, Pavel Gasnikov, Alexander Richtárik, Peter Mohamed bin Zayed University of Artificial Intelligence United Arab Emirates King Abdullah University of Science and Technology Saudi Arabia Moscow Institute of Physics and Technology Russia Université de Montréal and Mila Canada Canada CIFAR AI Canada Weierstrass Institute for Applied Analysis and Stochastics Germany University Innopolis Russia Ivannikov Institute for System Programming RAS Russia Skolkovo Institute of Science and Technology Russia

关键词： Stochastic systems

来源：评论

学校读者我要写书评

暂无评论

Mixed characteristic discontinuous Galerkin approach for perfect gas dynamics modeling

引用

Journal of Physics: Conference Series 2021年第1期2028卷

作者： V.K. Gatsuk V.V. Lukin Bauman Moscow State Technical University Moscow Russia Keldysh Institute of Applied Mathematics RAS Moscow Russia Ivannikov Institute for System Programming RAS Moscow Russia

The numerical method based on the discontinuous Galerkin (DG) approach for perfect inviscid compressible gas dynamics modelling is developed. The approximate Riemann solvers in combination with gas dynamics equations system characteristic properties are used to calculate the numerical fluxes on the cells interfaces and inside the cells. A number of approximate Riemann solvers are considered. The solution characteristic decomposition is used in the similar to PPML way. The developed algorithm is tested using problems with smooth and discontinuous solutions. Both one-dimensional and two-dimensional test problems are considered. The method provides accurate discontinuity resolution and ability to compute gas-dynamic instabilities with minimal artificial distortion.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Simulation and optimization of injector parameters with OpenFOAM and globalizer tools

引用

AIP Conference Proceedings 2023年第1期2849卷

作者： Konstantin Barkalov Ilya Lebedev Daniil Ryazanov Sergey Strijak Lobachevsky State University of Nizhny Novgorod Nizhny Novgorod Russia Ivannikov Institute for System Programming of the RAS Moscow Russia

The paper presents results of simulation and optimization of the processes taking place in a fuel system injector. Simula-tion has been carried using tools of the OpenFOAM system. The problem of differential pressure minimization at the injector inlet and outlet was set and solved. This problem was considered as a Lipschitz optimization problem with a black-box type objective function. To solve it, we used an efficient global search algorithm implemented in the Globalizer system.

关键词：

来源：评论

学校读者我要写书评

暂无评论

HIGHLY SMOOTH ZEROTH-ORDER METHODS FOR SOLVING OPTIMIZATION PROBLEMS UNDER THE PL CONDITION†

arXiv

引用

arXiv 2023年

作者： Gasnikov, A.V. Lobanov, A.V. Stonyakin, F.S. Moscow Institute of Physics and Technology 141700 Dolgoprudny Institutskiy per. 9 Russia Skolkovo Institute of Science and Technology 121205 Moscow B. Boulevard 30 bld. 1 Russia Institute for System Programming of the RAS 125047 Moscow A. Solzhenitsyn st. 25 Russia V.I. Vernadsky Crimean Federal University 295007 Simferopol Prospekt Vernadskogo 4 Russia

– In this paper, we study the black box optimization problem under the Polyak–Lojasiewicz (PL) condition, assuming that the objective function is not just smooth, but has higher smoothness. By using "kernel-based" approximations instead of the exact gradient in the Stochastic Gradient Descent method, we improve the best-known results of convergence in the class of gradient-free algorithms solving problems under the PL condition. We generalize our results to the case where a zeroth-order oracle returns a function value at a point with some adversarial noise. We verify our theoretical results on the example of solving a system of nonlinear equations. Copyright © 2023, The Authors. All rights reserved.

关键词： Gradient methods

来源：评论

学校读者我要写书评

暂无评论

Developing quasi-steady model for studying hemostatic response using supercomputer technologies

Supercomputing Frontiers and Innovations

引用

Supercomputing Frontiers and Innovations 2018年第4期5卷 67-72页

作者： Trifanov, Petr V. Kaneva, Valeriia N. Strijhak, Sergei V. Panteleev, Mikhail A. Ataullakhanov, Fazoil I. Dunster, Joanne Voevodin, Vadim V. Nechipurenko, Dmitry Y. Department of Physics Lomonosov Moscow State University Moscow Russia Dmitry Rogachev National Research Center of Pediatric Hematology Oncology and Immunology Moscow Russia Center for Theoretical Problems of Physicochemical Pharmacology Moscow Russia Ivannikov Institute for System Programming of the RAS Moscow Russia Faculty of Biological and Medical Physics Moscow Institute of Physics and Technology Dolgoprudny Russia Institute for Cardiovascular and Metabolic Research School of Biological Sciences University of Reading Berkshire United Kingdom Research Computing Center of Lomonosov Moscow State University Moscow Russia

Formation of the platelet plug represents a primary response to the vessel wall injury, but may also result in vessel occlusion. The decrease of the local blood flow due to platelet thrombus formation may lead to serious complications, such as ischemic stroke and myocardial infarction. However, mechanisms responsible for regulation of thrombus dynamics are not clear. In order to get a deeper insight into the role of blood flow and platelet interactions in the formation of the primary platelet plug we developed a particle-based model of microvascular thrombosis using quasisteady flow approximation. In order to simulate thrombus dynamics at physiologically relevant timescales of several minutes, we took advantage of the supercomputer technologies. Our in silico analysis revealed the importance of platelet size heterogeneity for describing experimental data on microvascular thrombus formation. Thus, our model represents a useful tool for the supercomputeraided computational analysis of thrombus dynamics in the microvessels on physiologically relevant timescales. © The Author 2018.

关键词： Platelets

来源：评论

学校读者我要写书评

暂无评论

Numerical simulation of compressible gas flows using regularized gas dynamic equations solver QGDFoam

引用

AIP Conference Proceedings 2018年第1期2027卷

作者： M. V. Kraposhin T. G. Elizarova M. A. Istomina D. A. Ryazanov K. A. Vatutin 1Ivannikov Institute for System Programming of RAS Moscow Russia 2Keldysh Institute of Applied Mathematics of RAS Moscow Russia 3Lomonosov Moscow State University Moscow Russia

The work presents results of the application of a new OpenFOAM® solver QGDFoam for the numerical simulation of viscous compressible flows in a wide range of Mach numbers. The new solver is based on the explicit approximation of regularized, or quasi-gas dynamic (QGD) equations. The mixed finite-volume and finite-difference approximation is constructed on unstructured spatial grids with co-located variables storage. The solver has been tested for a number of 1D Riemann problems (Sod’s problem, Noh test and others) and 2D cases (Mach 3 forward step, Ladenburg supersonic jet flow with Mach reflection, NASA Langley supersonic overexpanded jet flow and subsonic laminar flow over a backward-facing step). Results of numerical simulations were compared with analytic solutions and OpenFOAM® implementation of the Kurganov-Tadmor scheme, known as rhoCentralFoam. The testing procedure has shown that whereas QGD algorithm is more diffusive than Godunov-type methods with 2nd order TVD schemes with limiters, it is far less diffusive compared with pure upwind schemes as HLL. It was shown that OpenFOAM implementation of the QGD algorithm allows to compute successfully subsonic, sonic and supersonic flows, while other OpenFOAM® solvers have a very limited operational Mach number range. Preliminary results of QGDFoam application for large-scale 3D problems are presented. Scaling tests for up to 768 cores showed good scalability of QGDFoam solver.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Preface

引用

CEUR Workshop Proceedings 2017年 1989卷 5-7页

作者： Pozin, Boris A. Cavalli, Ana Rosa Petrenko, Alexander K. Department of Information and Analytical Systems EC-leasing Co. Russia Department of Network Software TELECOM and Management SudParis France Department of System Engineering Institute of System Programming of RAS Russia

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：