Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Laboratory of NumericaL
methods of
appLied
structuraL
optimizatioN
01.09.2018Laboratory NuMASO1
Mission &Goals
01.09.2018Laboratory NuMASO2
Mission Solving hard practical optimization problems State-of-the-art research in optimization theory and methods
Goals Find elegant and simple solutions for complex actual problems Create optimization method’s frameworks Extend the domain of optimization theory applicability Develop new numerical optimization methods and improve existing
ones Investigate optimization methods properties and effectiveness
Team. Lead Researchers
3
Dr. Mikhail PosypkinLead Researcher
Dr. Vitalii ZhadanLead Researcher
Dr. Alexandr GasnikovLead Researcher
Phd. Alexandr BiryukovSenior Researcher
Phd. Fedor StonyakinSenior Researcher
Dr. Vladimir ZubovLead Researcher
Phd. Andrey GorchakovSenior Researcher
Phd. Alexey ChernovSenior Researher – Deputy Head
Acad. Dr. Yury EvtushenkoChief Researcher – Head of the laboratory
Marina DanilovaJunior Researcher
Daniil MerkulovJunior Researcher
Dmitrii KamzolovJunior Researcher
01.09.2018Laboratory NuMASO
Laboratory Head Yury Evtushenko
01.09.2018Laboratory NuMASO4
Academic Degrees: 2006 Academician of the Russian
Academy of Sciences
1985 Professor
1981 Doctor of Physical and Mathematical Sciences, Computing Centre of the USSR Academy of Sciences (CCAS)
Research Interests: Linear and Nonlinear Programming
Decision Support Systems
Optimal Control
Optimization Techniques
Numerical Methods and Software for Solving Global MulticriteriaOptimization
Professional Experience: 1989 - p.t. Director of Dorodnicyn Computing
Centre of RAS 1981 - 1989 Deputy Director of Dorodnicyn
Computing Centre 1978 - 1981 Head of Subdepartment in the
Department of operation research, CCAS 1973 - 1978 Senior Researcher, CCAS 1967 - 1972 Junior Researcher, Computing Centre of
the USSR Academy of Sciences (CCAS) 1966 - 1967 Junior Researcher, Central
Aerohydrodynamic Institute (TsAGI) 1965 - 1966 Senior engineer (TsAGI) 1962 - 1965 Post graduate student at Moscow
Institute of Physics and Technology
Research Directions
Structural optimization Numerical methods of local and global optimization theory Multicriteria optimization Nonconvex and non-smooth optimization Linear and nonlinear programming, semidefinite and
equilibrium programming Numerical methods of optimal control Stochastic optimization High-performance techniques in optimization Distributed optimization
01.09.2018Laboratory NuMASO5
Competencies
01.09.2018Laboratory NuMASO6
Machine Learning Traffic assignment problems Computer networks modeling Inverse problems of mathematical physics Optimal Control in the Foundry Industry Global Optimization Robotics: Approximation of the Robot’s workspace Fast Automatic Differentiation Non-linear analysis and its applications Software Framework Software Development
Machine Learning
PageRank problem solution (nonconvex optimization)Cooperation with CORE UCL (Yu. Nesterov) Yandex
https://arxiv.org/pdf/1603.00717.pdf Distributed optimization with sum-type target functions
Cooperation with A. Nedichttps://arxiv.org/pdf/1712.00232.pdfhttps://arxiv.org/pdf/1803.02933.pdfhttps://arxiv.org/pdf/1806.03915.pdf
Image analysis (Monge-Kantorovich-Wasserstein transport distance)Cooperation with WIAS Berlin (P. Dvurechensky, V. Spokoiny)https://arxiv.org/pdf/1802.04367.pdf
01.09.2018Laboratory NuMASO7
Traffic assignment problems Applications
Cars transport (http://www.mathnet.ru/php/archive.phtml?wshow=paper&jrnid=crm&paperid=256&option_lang=rus)
Railways (https://arxiv.org/ftp/arxiv/papers/1501/1501.02205.pdf)
Short-term modeling and Long-term modeling: traffic assignment problems, mechanism design (road pricing) How to predict the situation on the roads for an hour
ahead? How to road price to achieve a system optimum? Where is it necessary to make reverse strips and dedicated
lanes for public transport? How to optimally manage traffic signaling and access to
major highways? How to optimally manage traffic flows?
Publics: Edited by Gasnikov A.V. Introduction in mathematical
modeling traffic flows // MCCME – Moscow, 2013.
01.09.2018Laboratory NuMASO8
Computer networks modeling & inverse problems of mathematical physics
Inverse problems of mathematical physicshttps://arxiv.org/ftp/arxiv/papers/1703/1703.00267.pdfCooperation with A. Kabanikhin (Novosibirsk)
Modeling of computer networks (Large scale and huge scale optimizationhttps://arxiv.org/ftp/arxiv/papers/1508/1508.00858.pdfCooperation with
01.09.2018Laboratory NuMASO9
Optimal Control in the Foundry Industry
01.09.2018Laboratory NuMASO10
Control of the process of crystallization of metal in the foundry industryA.V. Albu, A.F. Albu, V.I. Zubov. Mathematical modeling and control of the process of crystallization of metal in the foundry business. Scientific publication. Computing Center. A.A. Dorodnitsyn of the Russian Academy of Sciences. Moscow. 2012. 168 C.
Optimal control for foundry formu(t) is the velocity of the foundry form
Optimal Control in the Foundry Industry
01.09.2018Laboratory NuMASO11
Determination of the thermal conductivity of a substance by the dynamics of the temperature fieldA. Albu, V. Zubov. Identification of the thermal conductivity coefficient in two dimension case // Optim Lett (2018). https://doi.org/10.1007/s11590-018-1304-4 V. Zubov. Application of the methodology of rapid automatic differentiation to the solution of the inverse coefficient problem for the heat equation // Zh. Vychisl. Math. and Math. fiz. 2016. P. 56. № 10. P. 1760-1774.
"Recovery" of the function K (T) over a given temperature field with machine quality (accuracy)2100196.3 −⋅=iniФ 26109909.2 −⋅=optФ
Global Optimization
Overview
Publications
01.09.2018Laboratory NuMASO12
A deterministic global optimization method Allows to approximate function extrema, sets given by the system of equations, inequalities
or mapping with the pre-defined precision Method was successfully parallelized for multicore systems with shared and distributed
memory organization
1. Evtushenko Y. G. Computation of exact gradients in distributed dynamic systems // Optimization Methods and Software. – 1998. – V. 9. – P. 45-75
2. Evtushenko Y. G. Numerical methods for finding global extrema (case of a non-uniform mesh) //Computational Mathematics and Mathematical Physics. – 1971. – V. 11. – No. 6. – P. 38-54.
3. Evtushenko Y., Posypkin M. A deterministic approach to global box-constrained optimization //Optimization Letters. – 2013. – V. 7. – No. 4. – P. 819-829.
4. Evtushenko Y. G., Posypkin M. A. A deterministic algorithm for global multi-objective optimization //Optimization Methods and Software. – 2014. – V. 29. – No. 5. – P. 1005-1019.
5. Evtushenko Y., Posypkin M., Sigal I. A framework for parallel large-scale global optimization //Computer Science-Research and Development. – 2009. – V. 23. – No. 3-4. – P. 211-215.
6. Evtushenko, Y., Golubeva, Y., Orlov, Y., Posypkin, M.Using simulation for performance analysis and visualization of parallel Branch-and-Bound methods //Russian Supercomputing Days. – Springer, Cham, 2016. – P. 356-368.
7. Evtushenko Y. G., Posypkin M. A. Effective hull of a set and its approximation //Doklady Mathematics. –Pleiades Publishing, 2014. – V. 90. – No. 3. – P. 791-794.
8. Evtushenko, Y., Posypkin, M., Rybak, L., Turkin, A.. Approximating a solution set of nonlinear inequalities //Journal of Global Optimization. – 2017. – P. 1-17.
Robotics: Approximation of the Robot’s workspace
DexTAR parallel robot Workspace
01.09.2018Laboratory NuMASO13
Robotics: Approximation of the Robot’s workspace
The scientific reserve in the field of modeling of robotsof parallel structure is created.
Methods for approximating the working area of therobot and the corresponding software have beendeveloped.
Approaches have been developed for solving robotcontrol problems, based on the methodology of rapidautomatic differentiation of objective functions and taskconstraints.
01.09.2018Laboratory NuMASO14
Fast Automatic Differentiation
FAD technique: development of high-performance packages4D-Var data assimilations
Initial conditions search
spherical wave
01.09.2018Laboratory NuMASO15
Non-linear analysis and its applications
01.09.2018Laboratory NuMASO16
Applications of non-linear analysis inextremal problems (analogues of Ferma-Torrichelli-Shteiner problem)http://www.apmath.spbu.ru/cnsa/pdf/monograf/Выпуклый%20и%20негладкий%20анализ%20Орлов%20Стонякин%20Смирнова.pdf
Universal and adaptive methods for variationinequities and saddle problems. Equilibriumsearch problemshttps://arxiv.org/abs/1804.02579https://arxiv.org/abs/1806.05140
Optimization of functional with non-standard growth conditions, incl. arisingwhen solving problems Truss TopologyDesign.https://arxiv.org/abs/1710.06612https://arxiv.org/abs/1803.01329
Software framework
01.09.2018Laboratory NuMASO17
NUMERIC UTILITIES
INTERVAL ARITHMETICS
LIBRARY
EXPRESSIONS LIBRARY
NON-UNIFORM COVERING
METHOD “LOGIC”
SEQUENTIAL
SHARED MEMORY
DISTRIBUTED MEMORY
GPU (CUDA)
DISTRBUTED (BOINC)
Software framework: Expressions library
01.09.2018Laboratory NuMASO18
• Allows to declare complex formulas
• Supports automated evaluation of a function value, its derivatives, interval bounds and interval bounds for derivatives
template <classT> Expr<T> Rosenbrock(int n) {Expr<T> x;Iterator i(0, n-2);Expr<T> t = i;return loopSum(100*sqr(x[t+1]-sqr(x[i]))+sqr((x[i]-1)), i);
}
Software Development
Iterative and incremental Waterfall
01.09.2018Laboratory NuMASO19
Project Development
Life cycles
Starting the project
Organizing and Preparing
Carrying out the project
work
Closing the project
Software Development
01.09.2018Laboratory NuMASO20
Project Documentation
Version control systems
Software Development
01.09.2018Laboratory NuMASO21
Hadoop Technology stack
Programming Languages & Frameworks
Databases
Platforms
IDEs