Projection markovienne de processus stochastiques
Projection markovienne de processus stochastiques
Projection markovienne de processus stochastiques
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
tel-00766235, version 1 - 17 Dec 2012<br />
Curie- Paris VI) un<strong>de</strong>r the supervision of Rama Cont. It studies various<br />
mathematical aspects of problems related to the Markovian projection of<br />
stochastic processes, and explores some applications of the results obtained<br />
to mathematical finance, in the context of semimartingale mo<strong>de</strong>ls.<br />
Given a stochastic process ξ, mo<strong>de</strong>led as a semimartingale, our aim is<br />
to build a Markov process X whose marginal laws are the same as ξ. This<br />
constructionallowsustouseanalyticaltoolssuchasintegro-differentialequations<br />
to explore or compute quantities involving the marginal laws of ξ, even<br />
when ξ is not Markovian.<br />
We present a systematic study of this problem from probabilistic viewpoint<br />
and from the analytical viewpoint. On the probabilistic si<strong>de</strong>, given a<br />
discontinuous semimartingale we give an explicit construction of a Markov<br />
process X which mimics the marginal distributions of ξ, as the solution of a<br />
martingale problems for a certain integro-differential operator (Chapter 2).<br />
This construction extends the approach of Gyöngy to the discontinous case<br />
and applies to a wi<strong>de</strong> range of examples which arise in applications, in particular<br />
in mathematical finance. Some applications are given in Chapters 4<br />
and 5.<br />
On the analytical si<strong>de</strong>, we show that the flow of marginal distributions of<br />
a discontinuous semimartingale is the solution of an integro-differential equation,<br />
which extends the Kolmogorov forward equation to a non-Markovian<br />
setting. As an application, we <strong>de</strong>rive a forward equation for option prices<br />
in a pricing mo<strong>de</strong>l <strong>de</strong>scribed by a discontinuous semimartingale (Chapter 3).<br />
This forward equation generalizes the Dupire equation, originally <strong>de</strong>rived in<br />
the case of diffusion mo<strong>de</strong>ls, to the case of a discontinuous semimartingale.<br />
1.2.1 Chapter 2 : Markovian projection of semimartingales<br />
Consi<strong>de</strong>r, onafilteredprobabilityspace(Ω,F,(Ft)t≥0,P), anItosemimartingale,<br />
on the time interval [0,T], T > 0, given by the <strong>de</strong>composition<br />
ξt = ξ0+<br />
t<br />
0<br />
βsds+<br />
t<br />
0<br />
δsdWs+<br />
t<br />
0<br />
<br />
y≤1<br />
y ˜ M(dsdy)+<br />
t<br />
0<br />
<br />
y>1<br />
yM(dsdy),<br />
where ξ0 is in R d , W is a standard R n -valued Wiener process, M is an<br />
integer-valued random measure on [0,T]×R d with compensator measure µ<br />
12