parareal algorithm

Upload: tina-bargo

Post on 09-Apr-2018

217 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/8/2019 Parareal Algorithm

    1/43

    Parareal Algorithm

    Ma. Cristina Bargo

    Laboratoire Jacques-Louis Lions

    Universit Pierre et Marie Curie

    University of the Philippines Diliman

    CEMRACS 2009

    August 7, 2009

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 1 / 15

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    2/43

    Outline

    1 History

    2 The Algorithm

    3 Some Properties

    4 Work Done on Parareal Algorithm

    5 Simple Implementation

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 2 / 15

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    3/43

    History

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 3 / 15

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    4/43

    History

    Lions, Maday and Turinici [4] (2001) parallel in real time parareal

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 3 / 15

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    5/43

    History

    Lions, Maday and Turinici [4] (2001) parallel in real time parareal

    Bal and Maday [2] (2002) equivalent to [4] for linear problems better results for nonlinear problems

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 3 / 15

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    6/43

    History

    Lions, Maday and Turinici [4] (2001) parallel in real time parareal

    Bal and Maday [2] (2002) equivalent to [4] for linear problems better results for nonlinear problems

    Baffico et al [1] (2002) most common formulation

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 3 / 15

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    7/43

    The Problem

    Find u such that

    (1)

    tu+ A (t,u) = 0, t > t0

    u = u0, t = t0

    where A : R V V (V a Hilbert space) and t0 0.

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 4 / 15

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    8/43

    The Problem

    Find u such that

    (1)

    tu+ A (t,u) = 0, t > t0

    u = u0, t = t0

    where A : R V V (V a Hilbert space) and t0 0.

    u(t + ) = E(t + ,t,v), where v = u(t) and > 0, with

    > 0, > 0, E(t + + , t + , E(t + ,t,v)) = E(t + + ,t,v)

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 4 / 15

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    9/43

    The Problem

    Find u such that

    (1)

    tu+ A (t,u) = 0, t > t0

    u = u0, t = t0

    where A : R V V (V a Hilbert space) and t0 0.

    u(t + ) = E(t + ,t,v), where v = u(t) and > 0, with

    > 0, > 0, E(t + + , t + , E(t + ,t,v)) = E(t + + ,t,v)

    Let t0 = T0 < T1 < < TN = T, and Tn = Tn Tn1. Then

    n > 0, u(Tn) = E(Tn, T0,u0) = E(Tn, Tn1,u(Tn1))

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 4 / 15

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    10/43

    The Tools

    Notations :

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 5 / 15

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    11/43

    The Tools

    Notations :

    Fine solver F(t2, t1,u1)

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 5 / 15

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    12/43

    The Tools

    Notations :

    Fine solver F(t2, t1,u1) approximation of the solution u(t2) to problem (1) with initial

    condition u(t1) = u1

    can be a classical discretization scheme with small timestep t

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 5 / 15

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    13/43

    The Tools

    Notations :

    Fine solver F(t2, t1,u1) approximation of the solution u(t2) to problem (1) with initial

    condition u(t1) = u1

    can be a classical discretization scheme with small timestep tCoarse solver G(t2, t1,u1)

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 5 / 15

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    14/43

    The Tools

    Notations :

    Fine solver F(t2, t1,u1) approximation of the solution u(t2) to problem (1) with initial

    condition u(t1) = u1

    can be a classical discretization scheme with small timestep tCoarse solver G(t2, t1,u1) another approximation to u(t2), less accurate than F(t2, t1,u1) but

    cheaper to solve can be another discretization scheme with a larger timestep T

    (T >> t)

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 5 / 15

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    15/43

    The Algorithm

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 6 / 15

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    16/43

    The Algorithm

    Iteration 0 :

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 6 / 15

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    17/43

    The Algorithm

    Iteration 0 : U

    00 = u0 (the initial condition in problem (1))

    U0n+1 = G(tn+1, tn,U

    0n

    ) (the coarse solver)

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 6 / 15

    h l h

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    18/43

    The Algorithm

    Iteration 0 : U

    00 = u0 (the initial condition in problem (1))

    U0n+1 = G(tn+1, tn,U

    0n

    ) (the coarse solver)

    Iteration 1, 2,

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 6 / 15

    Th Al h

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    19/43

    The Algorithm

    Iteration 0 : U

    00 = u0 (the initial condition in problem (1))

    U0n+1 = G(tn+1, tn,U

    0n

    ) (the coarse solver)

    Iteration 1, 2, Parallel : F(tn+1, tn,Ukn) for n = 0, 1, 2, , n 1

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 6 / 15

    Th Al i h

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    20/43

    The Algorithm

    Iteration 0 : U

    00 = u0 (the initial condition in problem (1))

    U0n+1 = G(tn+1, tn,U

    0n

    ) (the coarse solver)

    Iteration 1, 2, Parallel : F(tn+1, tn,Ukn) for n = 0, 1, 2, , n 1 Serial : Uk0 = u0 and

    (2) Uk+1n+1 = G(tn+1, tn,U

    k+1n

    ) + F(tn+1, tn,Uk

    n) G(tn+1, tn,U

    k

    n)

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 6 / 15

    S P i

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    21/43

    Some Properties

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 7 / 15

    S P ti

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    22/43

    Some Properties

    For k , then Ukn Un where

    Un+1 = F(tn+1, t0,U0) = F(tn+1, tn,Un)

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 7 / 15

    S P ti

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    23/43

    Some Properties

    For k , then Ukn Un where

    Un+1 = F(tn+1, t0,U0) = F(tn+1, tn,Un)

    n = 0, 1, 2, , N we can show that Unn = Un

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 7 / 15

    Some Properties

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    24/43

    Some Properties

    For k , then Ukn Un where

    Un+1 = F(tn+1, t0,U0) = F(tn+1, tn,Un)

    n = 0, 1, 2, , N we can show that Unn = Un

    converges much faster (Maday, Rnquist and Staff, 2006 [5])

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 7 / 15

    Some Properties

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    25/43

    Some Properties

    For k , then Ukn Un where

    Un+1 = F(tn+1, t0,U0) = F(tn+1, tn,Un)

    n = 0, 1, 2, , N we can show that Unn = Un

    converges much faster (Maday, Rnquist and Staff, 2006 [5])

    stability results (Staff and Rnquist, 2005 [8])

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 7 / 15

    Work Done on Parareal Algorithm

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    26/43

    Work Done on Parareal Algorithm

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 8 / 15

    Work Done on Parareal Algorithm

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    27/43

    Work Done on Parareal Algorithm

    pricing of an American put (Bal and Maday, 2002 [2])

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 8 / 15

    Work Done on Parareal Algorithm

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    28/43

    Work Done on Parareal Algorithm

    pricing of an American put (Bal and Maday, 2002 [2])molecular dynamics simulations (Baffico et al, 2002 [1]) - fine schemeuses the full model, coarse scheme is based on a simpler model of theoriginal

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 8 / 15

    Work Done on Parareal Algorithm

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    29/43

    Work Done on Parareal Algorithm

    pricing of an American put (Bal and Maday, 2002 [2])molecular dynamics simulations (Baffico et al, 2002 [1]) - fine schemeuses the full model, coarse scheme is based on a simpler model of theoriginal

    control problems (Maday and Turinici, 2002 [6])

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 8 / 15

    Work Done on Parareal Algorithm

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    30/43

    Work Done on Parareal Algorithm

    pricing of an American put (Bal and Maday, 2002 [2])molecular dynamics simulations (Baffico et al, 2002 [1]) - fine schemeuses the full model, coarse scheme is based on a simpler model of theoriginal

    control problems (Maday and Turinici, 2002 [6])combined with domain-decomposition methods (Maday and Turinici,2005 [7])

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 8 / 15

    Work Done on Parareal Algorithm

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    31/43

    Work Done on Parareal Algorithm

    pricing of an American put (Bal and Maday, 2002 [2])molecular dynamics simulations (Baffico et al, 2002 [1]) - fine schemeuses the full model, coarse scheme is based on a simpler model of theoriginal

    control problems (Maday and Turinici, 2002 [6])combined with domain-decomposition methods (Maday and Turinici,2005 [7])

    Navier-Stokes (Fischer, Hecht and Maday, 2005 [3]) - using fine andcoarse mesh in space

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 8 / 15

    Work Done on Parareal Algorithm

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    32/43

    Work Done on Parareal Algorithm

    pricing of an American put (Bal and Maday, 2002 [2])molecular dynamics simulations (Baffico et al, 2002 [1]) - fine schemeuses the full model, coarse scheme is based on a simpler model of theoriginal

    control problems (Maday and Turinici, 2002 [6])combined with domain-decomposition methods (Maday and Turinici,2005 [7])

    Navier-Stokes (Fischer, Hecht and Maday, 2005 [3]) - using fine andcoarse mesh in space

    Goal : Use the algorithm to solve more complicated problems

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 8 / 15

    Simple Implementation

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    33/43

    S p e p e e tat o

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 9 / 15

    Simple Implementation

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    34/43

    p p

    Sample Problem : Find u =

    u1 u2T

    so that

    u

    t= Au, t (0, 100]

    u = u0 = 1 0 T

    , t = 0

    where A is the 2 2 matrix given by

    A =

    0.02 0.20.2 0.02

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 9 / 15

    Simple Implementation

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    35/43

    p p

    t0 = 0 and T = 100

    N = 100 and Tn = T = 1

    Coarse scheme : Implicit Euler with T = T = 1

    Fine scheme : Implicit Euler with t = 0.1Horizontal axis is u1 and the vertical axis is u2

    fine scheme : 0.62 s

    parareal scheme : 0.29496 s

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 10 / 15

    Simple Implementation

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    36/43

    p pIteration 0

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 11 / 15

    Simple Implementation

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    37/43

    Iteration 1

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 12 / 15

    Simple Implementation

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    38/43

    Iteration 2

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 13 / 15

    Simple Implementation

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    39/43

    Iteration 3

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 14 / 15

    Simple Implementation

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    40/43

    Iteration 4

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 15 / 15

    L. Baffico, S. Bernard, Y. Maday, G. Turinici, and G. Zrah.Parallel in time molecular dynamics simulations

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    41/43

    Parallel-in-time molecular-dynamics simulations.Phys. Rev. E, 66(5) :057701, Nov 2002.

    Guillaume Bal and Yvon Maday.

    A parareal time discretization for nonlinear PDEs with application tothe pricing of an american put.In L.F. Pavarino and A. Toselli, editors, Recent Developments inDomain Decomposition Methods, volume 23 of Lecture Notes inComputational Science and Engineering

    , pages 189202.Springer-Verlag, Berlin, 2002.

    Paul F. Fischer, Frdric Hecht, and Yvon Maday.A parareal in time semi-implicit approximation of the Navier-Stokesequations.

    In Domain Decomposition Methods in Science and Engineering,volume 40 of Lecture Notes in Computational Science and Engineering,pages 433440. Springer, Berlin, 2005.

    Jacques-Louis Lions, Yvon Maday, and Gabriel Turinici.

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 15 / 15

    Rsolution dEDP par un schma en temps pararel.C. R. Acad. Sci. Paris Sr. I Math., 332(7) :661668, 2001.

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    42/43

    Y. Maday, E.M. R nquist, and G.A. Staff.The parareal-in-time algorithm : basics, stability and more.

    2006.Yvon Maday and Gabriel Turinici.A parareal in time procedure for the control of partial differentialequations.C. R. Math. Acad. Sci. Paris, 335(4) :387392, 2002.

    Yvon Maday and Gabriel Turinici.The parareal in time iterative solver : a further direction to parallelimplementation.In Domain decomposition methods in science and engineering,

    volume 40 of Lecture Notes in Computational Science and Engineering,pages 441448. Springer, Berlin, 2005.

    Gunnar Andreas Staff and Einar M. Rnquist.Stability of the parareal algorithm.

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 15 / 15

    In Domain Decomposition Methods in Science and Engineering,volume 40 of Lecture Notes in Computational Science and Engineering,

    http://goforward/http://find/http://goback/
  • 8/8/2019 Parareal Algorithm

    43/43

    pages 449456. Springer, Berlin, 2005.

    MC Bargo (UPMC + UPD) Parareal Algorithm CEMRACS 2009 15 / 15

    http://goforward/http://find/http://goback/