16.08.2013 Views

Projection markovienne de processus stochastiques

Projection markovienne de processus stochastiques

Projection markovienne de processus stochastiques

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

tel-00766235, version 1 - 17 Dec 2012<br />

<strong>de</strong>fined, for f ∈ C ∞ 0 (Rd ), by<br />

Ltf(x) = b(t,x).∇f(x)+<br />

<br />

+<br />

R d<br />

d<br />

i,j=1<br />

aij(t,x)<br />

2<br />

∂2f (x)<br />

∂xi∂xj<br />

[f(x+y)−f(x)−1{y≤1}y.∇f(x)]n(t,dy,x),<br />

(2.3)<br />

where a : [0,T]×R d ↦→ Md×d(R), b : [0,T]×R d ↦→ R d and<br />

n : [0,T]×R d ↦→ R(R d −{0}) are measurable functions.<br />

For (t0,x0) ∈ [0,T]×R d , we recall that a probability measure Qt0,x0 on<br />

(Ω0,BT) is a solution to the martingale problem for (L,C ∞ 0 (Rd )) on [0,T] if<br />

Q(Xu = x0, 0 ≤ u ≤ t0) = 1 and for any f ∈ C ∞ 0 (R d ), the process<br />

f(Xt)−f(x0)−<br />

t<br />

t0<br />

Lsf(Xs)ds<br />

isa(Qt0,x0,(Bt))-martingaleon[0,T]. Existence, uniqueness andregularityof<br />

solutions to martingale problems for integro-differential operators have been<br />

studied un<strong>de</strong>r various conditions on the coefficients [90, 59, 40, 66, 77, 42].<br />

We make the following assumptions on the coefficients:<br />

Assumption 2.1 (Boun<strong>de</strong>dness of coefficients).<br />

(i) ∃K1 > 0, ∀(t,z) ∈ [0,T]×R d <br />

1∧y 2<br />

, b(t,z)+a(t,z)+<br />

n(t,dy,z) ≤ K1<br />

(ii) lim<br />

R→∞<br />

T<br />

sup<br />

0 z∈Rd n(t,{y ≥ R},z) dt = 0.<br />

where . <strong>de</strong>notes the Eucli<strong>de</strong>an norm.<br />

Assumption 2.2 (Continuity).<br />

(i) For t ∈ [0,T] and B ∈ B(R d − {0}), b(t,.), a(t,.) and n(t,B,.) are<br />

continuous on R d , uniformly in t ∈ [0,T].<br />

(ii) For all z ∈ R d , b(.,z), a(.,z) and n(.,B,z) are continuous on [0,T[,<br />

uniformly in z ∈ R d .<br />

28

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!