Mathematisch-
Naturwissenschaftliche Fakultät
Fachbereich Mathematik
Prof. Dr. Andreas Prohl Dr. Ananta Kumar Majee
Stochastische Differentialgleichungen
Summer Semester 2017 Tübingen, 26.04.2017
Homework 2
Problem 1.LetT >0, and(Ω,F,P)be a probability space. In the lecture, we introduced the Lagrange interpolation
Yek,h(t); 0≤t≤T via
Yek,h(t) = tj+1−t
k Yk,h(tj) +t−tj
k Yk,h(tj+1) ∀tj ≤t < tj+1, whereYk,h(tj) =Pj
l=1ηlis the random walk, andtj =jTJ =:jkare the mesh points of the equi-distant meshGk :=
tj Jj=0 covering[0, T]. Letf ∈Cb(R). Show that the iterates
uj Jj=0 ⊂ Cb(R)be given by
uj(x) =E h
f x+Yek,√k(tj)i
∀x∈R solves
(uj(x)−uj−1(x)
k = uj−1(x+
√k)−2uj−1(x)+uj−1(x−√ k)
2k (1≤j≤J)
u0(x) =f(x) ∀x∈R, (1)
where (1) is a finite difference discretization of the linear heat equation.
Problem 2.Lett >0. In the lecture, we define the quadratic variation of a Wiener processW QW(t) =L2−lim
J↑∞QWkJ(t), where QWkJ(t) = PJ
j=1
W(tJj)− W(tJj−1)
2 on the equi-distant mesh of size k > 0 covering [0, t]
IkJ =
0, tJ1,· · · , tJJ wheretJj = jtJ =:jk. Let
Ikl;l∈N be a sequence of equi-distant meshes such thatP∞
l=1kl<∞. Show that
QWkl(t)→t P-a.s. (l↑ ∞)
Hint:Use Tschebycheff’s inequality and Borel-Cantelli lemma to transfer the corresponding L2-convergence result from the lecture.
Problem 3. Let0 = tJ0 < tJ1 < · · · < tJJ = T, wheretJj = jTJ be an equi-distant mesh of sizek = TJ covering[0, T]. Find the following limits
a). L2−lim
J↑∞
J−1
X
j=0
W(tJj)h
W(tJj+1)−W(tJj)i ,
b). L2− lim
J↑∞
J−1
X
j=0
W(tJj+1)h
W(tJj+1)−W(tJj)i .
Seite 1/2
Problem 4.Letf ∈Mstep2 such thatf(t) =PJ−1
j=0 ηj1[tj,tj+1)(t). Consider the stochastic integral
I(f) =
J−1
X
j=1
ηj
W(tj+1)−W(tj) .
Show that
E h
I(f)
2i
=E hZ ∞
0
|f(s)|2dsi .
Date of submission: 03.05.2017.
Seite 2/2