Logout succeed
Logout succeed. See you again!

Taylor Expansions and Numerical Approximations for Stochastic Partial Differential Equations PDF
Preview Taylor Expansions and Numerical Approximations for Stochastic Partial Differential Equations
TaylorexpansionsforSODEs TaylorexpansionsforSPDEs AnewnumericalmethodforSPDEswithnon-additivenoise AnewnumericalmethodforSPDEswithadditivenoise Taylor Expansionsand Numerical Approximationsfor Stochastic Partial Differential Equations A.Jentzen JointworkswithP.E.KloedenandM.Röckner FacultyofMathematics BielefeldUniversity 12thAugust2010 A.Jentzen TaylorExpansionsforSPDEs TaylorexpansionsforSODEs TaylorexpansionsforSPDEs AnewnumericalmethodforSPDEswithnon-additivenoise AnewnumericalmethodforSPDEswithadditivenoise Content 1 TaylorexpansionsforSODEs 2 TaylorexpansionsforSPDEs 3 AnewnumericalmethodforSPDEswithnon-additivenoise 4 AnewnumericalmethodforSPDEswithadditivenoise A.Jentzen TaylorExpansionsforSPDEs TaylorexpansionsforSODEs TaylorexpansionsforSPDEs AnewnumericalmethodforSPDEswithnon-additivenoise AnewnumericalmethodforSPDEswithadditivenoise Content 1 TaylorexpansionsforSODEs 2 TaylorexpansionsforSPDEs 3 AnewnumericalmethodforSPDEswithnon-additivenoise 4 AnewnumericalmethodforSPDEswithadditivenoise A.Jentzen TaylorExpansionsforSPDEs TaylorexpansionsforSODEs TaylorexpansionsforSPDEs AnewnumericalmethodforSPDEswithnon-additivenoise AnewnumericalmethodforSPDEswithadditivenoise LetT > 0andlet(Ω,F,P)beaprobabilityspace.Letf,g : R→ Rbe smoothfunctionsandlet(W) beascalarBrownianmotion. t t∈[0,T] ConsidertheSODE: dX = f(X )dt +g(X )dW, t t t t whichisunderstoodas t t X = X + f(X )ds+ g(X )dW t 0 s s s Z0 Z0 P-a.s.forallt ∈ [0,T].ApplyingItˆo’sformulatotheintegrandsaboveyields t t s X ≈ X +f(X )·t +g(X )· dW +g′(X )g(X )· dW dW t 0 0 0 s 0 0 u s Z0 Z0 Z0 1 = X +f(X )·t +g(X )·W + ·g′(X )g(X )· (W)2−t P-a.s. 0 0 0 t 0 0 t 2 (cid:0) (cid:1) Milstein’sapproximation(1974). A.Jentzen TaylorExpansionsforSPDEs TaylorexpansionsforSODEs TaylorexpansionsforSPDEs AnewnumericalmethodforSPDEswithnon-additivenoise AnewnumericalmethodforSPDEswithadditivenoise LetT > 0andlet(Ω,F,P)beaprobabilityspace.Letf,g : R→ Rbe smoothfunctionsandlet(W) beascalarBrownianmotion. t t∈[0,T] ConsidertheSODE: dX = f(X )dt +g(X )dW, t t t t whichisunderstoodas t t X = X + f(X )ds+ g(X )dW t 0 s s s Z0 Z0 P-a.s.forallt ∈ [0,T].ApplyingItˆo’sformulatotheintegrandsaboveyields t t s X ≈ X +f(X )·t +g(X )· dW +g′(X )g(X )· dW dW t 0 0 0 s 0 0 u s Z0 Z0 Z0 1 = X +f(X )·t +g(X )·W + ·g′(X )g(X )· (W)2−t P-a.s. 0 0 0 t 0 0 t 2 (cid:0) (cid:1) Milstein’sapproximation(1974). A.Jentzen TaylorExpansionsforSPDEs TaylorexpansionsforSODEs TaylorexpansionsforSPDEs AnewnumericalmethodforSPDEswithnon-additivenoise AnewnumericalmethodforSPDEswithadditivenoise LetT > 0andlet(Ω,F,P)beaprobabilityspace.Letf,g : R→ Rbe smoothfunctionsandlet(W) beascalarBrownianmotion. t t∈[0,T] ConsidertheSODE: dX = f(X )dt +g(X )dW, t t t t whichisunderstoodas t t X = X + f(X )ds+ g(X )dW t 0 s s s Z0 Z0 P-a.s.forallt ∈ [0,T].ApplyingItˆo’sformulatotheintegrandsaboveyields t t s X ≈ X +f(X )·t +g(X )· dW +g′(X )g(X )· dW dW t 0 0 0 s 0 0 u s Z0 Z0 Z0 1 = X +f(X )·t +g(X )·W + ·g′(X )g(X )· (W)2−t P-a.s. 0 0 0 t 0 0 t 2 (cid:0) (cid:1) Milstein’sapproximation(1974). A.Jentzen TaylorExpansionsforSPDEs TaylorexpansionsforSODEs TaylorexpansionsforSPDEs AnewnumericalmethodforSPDEswithnon-additivenoise AnewnumericalmethodforSPDEswithadditivenoise LetT > 0andlet(Ω,F,P)beaprobabilityspace.Letf,g : R→ Rbe smoothfunctionsandlet(W) beascalarBrownianmotion. t t∈[0,T] ConsidertheSODE: dX = f(X )dt +g(X )dW, t t t t whichisunderstoodas t t X = X + f(X )ds+ g(X )dW t 0 s s s Z0 Z0 P-a.s.forallt ∈ [0,T].ApplyingItˆo’sformulatotheintegrandsaboveyields t t s X ≈ X +f(X )·t +g(X )· dW +g′(X )g(X )· dW dW t 0 0 0 s 0 0 u s Z0 Z0 Z0 1 = X +f(X )·t +g(X )·W + ·g′(X )g(X )· (W)2−t P-a.s. 0 0 0 t 0 0 t 2 (cid:0) (cid:1) Milstein’sapproximation(1974). A.Jentzen TaylorExpansionsforSPDEs TaylorexpansionsforSODEs TaylorexpansionsforSPDEs AnewnumericalmethodforSPDEswithnon-additivenoise AnewnumericalmethodforSPDEswithadditivenoise LetT > 0andlet(Ω,F,P)beaprobabilityspace.Letf,g : R→ Rbe smoothfunctionsandlet(W) beascalarBrownianmotion. t t∈[0,T] ConsidertheSODE: dX = f(X )dt +g(X )dW, t t t t whichisunderstoodas t t X = X + f(X )ds+ g(X )dW t 0 s s s Z0 Z0 P-a.s.forallt ∈ [0,T].ApplyingItˆo’sformulatotheintegrandsaboveyields t t s X ≈ X +f(X )·t +g(X )· dW +g′(X )g(X )· dW dW t 0 0 0 s 0 0 u s Z0 Z0 Z0 1 = X +f(X )·t +g(X )·W + ·g′(X )g(X )· (W)2−t P-a.s. 0 0 0 t 0 0 t 2 (cid:0) (cid:1) Milstein’sapproximation(1974). A.Jentzen TaylorExpansionsforSPDEs TaylorexpansionsforSODEs TaylorexpansionsforSPDEs AnewnumericalmethodforSPDEswithnon-additivenoise AnewnumericalmethodforSPDEswithadditivenoise LetT > 0andlet(Ω,F,P)beaprobabilityspace.Letf,g : R→ Rbe smoothfunctionsandlet(W) beascalarBrownianmotion. t t∈[0,T] ConsidertheSODE: dX = f(X )dt +g(X )dW, t t t t whichisunderstoodas t t X = X + f(X )ds+ g(X )dW t 0 s s s Z0 Z0 P-a.s.forallt ∈ [0,T].ApplyingItˆo’sformulatotheintegrandsaboveyields t t s X ≈ X +f(X )·t +g(X )· dW +g′(X )g(X )· dW dW t 0 0 0 s 0 0 u s Z0 Z0 Z0 1 = X +f(X )·t +g(X )·W + ·g′(X )g(X )· (W)2−t P-a.s. 0 0 0 t 0 0 t 2 (cid:0) (cid:1) Milstein’sapproximation(1974). A.Jentzen TaylorExpansionsforSPDEs TaylorexpansionsforSODEs TaylorexpansionsforSPDEs AnewnumericalmethodforSPDEswithnon-additivenoise AnewnumericalmethodforSPDEswithadditivenoise LetT > 0andlet(Ω,F,P)beaprobabilityspace.Letf,g : R→ Rbe smoothfunctionsandlet(W) beascalarBrownianmotion. t t∈[0,T] ConsidertheSODE: dX = f(X )dt +g(X )dW, t t t t whichisunderstoodas t t X = X + f(X )ds+ g(X )dW t 0 s s s Z0 Z0 P-a.s.forallt ∈ [0,T].ApplyingItˆo’sformulatotheintegrandsaboveyields t t s X ≈ X +f(X )·t +g(X )· dW +g′(X )g(X )· dW dW t 0 0 0 s 0 0 u s Z0 Z0 Z0 1 = X +f(X )·t +g(X )·W + ·g′(X )g(X )· (W)2−t P-a.s. 0 0 0 t 0 0 t 2 (cid:0) (cid:1) Milstein’sapproximation(1974). A.Jentzen TaylorExpansionsforSPDEs