Thanks to visit codestin.com
Credit goes to www.scribd.com

0% found this document useful (0 votes)
85 views44 pages

Stochastic Diff Eq PDF

This document introduces Ito integrals and stochastic processes. It defines a stochastic process as a collection of random variables indexed by time. Brownian motion and the Wiener process are introduced as important examples of stochastic processes. The document discusses how the derivative of Brownian motion can be represented as white noise. This leads to defining the Ito integral, which provides a way to make mathematical sense of stochastic integrals involving white noise terms. The Ito integral allows modeling stochastic differential equations driven by white noise.
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
85 views44 pages

Stochastic Diff Eq PDF

This document introduces Ito integrals and stochastic processes. It defines a stochastic process as a collection of random variables indexed by time. Brownian motion and the Wiener process are introduced as important examples of stochastic processes. The document discusses how the derivative of Brownian motion can be represented as white noise. This leads to defining the Ito integral, which provides a way to make mathematical sense of stochastic integrals involving white noise terms. The Ito integral allows modeling stochastic differential equations driven by white noise.
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 44

lto lntcra-

Henrik Madsen
January 2006
Contents
1 Stochastic processes 2
2 Ito Integrals 9
21 Thc lto intcra 20
211 Somc jrojcrtic- ot thc lto intcra 2
212 Lxtcn-ion- ot thc lto intcra 3!
213 `uti-dimcn-iona lto intcra 3!
21! lto intcra dcncd y convcrcncc in
jroaiity 3o
21` A comjari-on ot lto and Stratonovich
intcra- 3
21o Thc lto jrocc-- and lto tormua !1
1
Chapter 1
Stochastic processes
Denition 1.1 A stochastic process is a parametrized
collection of random variables
{X
t
}
tT
dened on a probability space (, F, P) and assuming
values in R
n
.
Thc jaramctcr -jacc T i- u-uay thc hat-inc |0, ), indi-
catin thc timc Ior cach t T xcd wc havc a random
variac
X
t
().
On thc othcr hand, xin wc can con-idcr thc tunction
t X
t
(). t T
which i- cacd thc path ot X
t

So rouhy -jcakin a -tocha-tic jrocc-- i- a -y-tcm which


dcvcoj in timc in accordancc with jroaii-tic aw-
2
Example 1.1 Figure 1.1 illustrates one of the simplest
stochastic processes. Here X
n
is a random variable de-
noting the position at time n of a moving particle (n
0, 1, 2, . . .). Initially the particle is at the origin, X
0
0.
At n 1, there is a jump of one step, upwards to position
1 with probability
1
2
, and downwards to position -1 with
probability
1
2
. At n 2 there is a further jump, again
of one step upwards or downwards with equal probabil-
ity, the jumps at times n 1, 2 being independent. In
general,
X
n
X
n1
+ Z
n
(11)
where Z
n
, the jump at the nth step, is such that the ran-
dom variables {Z
1
, Z
2
, . . .} are mutually Bernoulli ran-
dom variables with thedistribution
Prob(Z
n
1) Prob(Z
n
1)
1
2
(n 1, 2, . . .).
(12)
3
Equation 1.1 taken with the initial condition X
0
0, is
equivalent to
X
n
Z
1
+ + Z
n
. (13)
A stochastic process on the form 1.3 is called a random walk.
O
O
O
O
O
O
O
O
O
O
O
O
O
time
P
o
s
i
t
i
o
n
0 2 4 6 8 10 12
-
2
0
2
4
Two independent realizations of a simple random walk
X
X
X
X
X
X
X
X
X
X
X
X
X
Figure 1.1:
4
An imjortant cncraization ot thc random wak i- whcn thc
,umj- may hajjcn at any timc t and whcn thc -izc ot thc
,umj- toow thc norma di-triution Thi- jrocc-- actuay
ajjcar- y a -iht modication ot thc jrocc-- aovc
Example 1.2 Suppose independent steps of length

t
are taken at time intervals of length t, with
Z
i

_
+1 for a positive step
1 for a negative step
where again Z
i
are independent Bernoulli random vari-
ables.
Now we dene
Y
t

t(Z
1
+ + Z
[t/t]
) (1!)
where |.| denotes the integer part. Then
E|Y
t
| 0, V |Y
t
| t|t/t| t
as t 0, and furthermore the central limit theorem
implies that Y
t
W
t
, where W
t
is gaussian. Actually
W
t
is a \icncr jrocc-- (Gard, 1988).
5
Denition 1.2 A Wiener process beginning at 0, with
drift and variance is a process W
t
with following
properties:
1. P{W
0
0} 1.
2. For all non-overlapping time intervals |t
1
, t
2
|, |t
3
, t
4
|
the random variables W
t
2
W
t
1
and W
t
4
W
t
3
are
independent.
3. For any time interval |t
1
, t
2
| W
t
2
W
t
1
is gaussian
distributed with
E|W
t
2
W
t
1
| (t
2
t
1
), (1`)
V |W
t
2
W
t
1
|
2
(t
2
t
1
). (1o)
lt 0 ic, thc mcan i- 0 thcn thc jrocc-- i- cacd
Brownian motion, dcnotcd B
t
lt a-o
2
1 thc thc
jrocc-- i- cacd standard Brownian motion -incc thcn
W
t
2
W
t
1
N(0, 1)
6
\c now -tatc -omc imjortant jrojcrtic- ot Lrownian motion
1 E|B
t
| 0
2 V |B
t
| E|B
t
B
t
|
2
t
3 cov|B
t+s
, B
t
| cov|B
t
, B
t+s
| E|B
t
B
t+s
|
2
t
! Thc jath ot B
t
() (ic thc tunction t B
t
()) i-
continuou- tor amo-t a
Proof.
1 E|B
t
| E|B
t
B
0
+ B
0
| E|B
t
B
0
| + E|B
0
|
0 + 0 0.
2 V |B
t
| V |B
t
B
0
+ B
0
| V |B
t
B
0
| + V |B
0
|

2
t + 0
2
t
3 cov|B
t+s
, B
t
| cov|B
t+s
B
t
+ B
t
, B
t
| cov|B
t+s

B
t
, B
t
| + cov|B
t
, B
t
| cov|B
t+s
B
t
, B
t
B
0
+B
0
| +
V |B
t
| cov|B
t+s
B
t
, B
t
B
0
| +cov|B
t+s
B
t
, B
0
| +
V |B
t
| 0 + 0 +
2
t
! Ior a jroot ot thi- -cc |1!, Ok-cnda| Loo-cy -jcakin
it i- jrovcn that B
t
() can c cxtcndcd to a continuou-
jrocc-- Y
t
() whcrc P(Y
t
() B
t
()) 1, tor a t
Thc jrocc--c- Y
t
() and B
t
() miht dicr cvcn thouh
thc jroaiity ot cquaity i- 1, thi- cxjain- why wc -ay
that thc jath i- continuou- tor amo-t a
7
Denition 1.3 A stochastic process X
t
is called stationary
if {X
t
} has the same distribution as {X
t+h
} for any
h > 0.
`otc that thc Lrownian motion i- not -tationary -incc V |B
t
|
V |B
s
| it s t. ut the increment B
t+h
B
t
is stationary
Iurthcrmorc, wc havc that
Thc Lrownian motion B
t
i- continuou- (`orc corrccty
B
t
ha- a continuou- version, and wc a--umc that wc
con-idcr -uch a continuou- vcr-ion)
Thc Lrownian motion B
t
i- nowhcrc dicrcntiac
Proof.
Thc non-dicrcntiaiity ot thc jrocc--, c in mcan
-quarc -cn-c, can c -ccn trom
E
_
(B
t+h
B
t
)
2
h
_


2
h
h
2


2
h
. h 0.
lt not -tatc othcrwi-c wc a--umc in thc toowin that B
t
i-
a standard Brownian motion
8
Chapter 2
Ito Integrals
ln thc toowin wc tocu- on ndin a rca-onac mathc-
matica intcrjrctation ot thc noi-c tcrm in thc cquation
da(t)
dt
(r(t) +

noise

)a(t)
ln a morc cncra torm
dX
dt
b(t, X
t
) + (t, X
t
)

noise

(21)
whcrc b and arc -omc ivcn tunction-
lt i- rca-onac to ook tor -omc -tocha-tic jrocc-- W
t
to
rcjrc-cnt thc noi-c tcrm, -o that
dX
dt
b(t, X
t
) + (t, X
t
)W
t
. (22)
La-cd on many -ituation-, onc i- cd to a--umc that W
t
ha-,
at ca-t ajjroximatcy thc-c jrojcrtic-
1 t
1
t
2
W
t
1
and W
t
2
arc indcjcndcnt,
2 W
t
i- -tationary,
3 E|W
t
| 0 tor a t
9
lt turn- out that thcrc doc- not cxi-t any rca-onac
-tocha-tic jrocc-- -ati-tyin oth 1 and 2, -uch a W
t
cannot
havc continuou- jath
10
Lowcvcr, thc covariancc tunction ot thc Lrownian motion i-
(s, t) min(s, t)
Lcncc
(s, t)
s

_
1 s < t
0 s > t
Thi- imjic- that thc mixcd -ccond dcrivativc doc- not cxi-t
Lowcvcr, it wc cacuatcd tormay

2
(s, t)
st
(s t)
That i-, tormay thc covariancc tunction ot thc dcrivativc ot
thc \icncr jrocc-- (or Lrownian motion) i- thc covariancc
tunction ot whitc noi-c, and hcncc that thc dcrivativc ot B
t
i- white noise
Iormay, W
t
aovc corrc-jond- to thc dcrivativc ot thc B
t

Thi- i- thc ackround tor rcjrc-cntin W


t
a- a cncraizcd
-toca-tic jrocc-- cacd thc white noise process
11
A morc rca-onac con-truction i- otaincd y rcwritin
cquation 22
`ow ct 0 t
0
< t
1
< < t
m
t and con-idcr a di-crctc
vcr-ion ot 22
X
k+1
X
k
b(t
k
, X
k
)t
k
+ (t
k
, X
k
)W
k
t
k
(23)
whcrc
X
j
X
t
j
, W
k
W
t
k
, t
k
t
k+1
t
k
.
`ow rcjacc W
k
t
k
y V
k
V
k+1
V
k
whcrc {V
t
}
t0
i-
-omc -uitac -tocha-tic jrocc--
Thc a--umjtion- 1,2, and 3 on W
t
-uc-t that V
t
-houd
havc stationary independent increments with mean 0
Such a jrocc-- cxi-t- and it turn- out that thc ony -uch
jrocc-- with continuou- jath- i- thc Lrownian motion B
t

(`otc that cvcn thouh thc Lrownian motion i- not -tation-


ary thc incrcmcnt- arc)
Thu- wc jut V
t
B
t
and otain trom cquation 23
X
k
X
0
+
k1

j=0
b(t
k
, X
k
)t
k
+
k1

j=0
(t
k
, X
k
)B
j
. (2!)
12
lt it i- jo--ic to jrovc that thc imit ot thc riht hand -idc
ot cquation 2! cxi-t- whcn t
j
0, thcn y ajjyin thc
u-ua intcration notation wc -houd otain
X
k
X
0
+
_
t
0
b(s, X
s
)ds +

_
t
0
(s, X
s
)dB

s
. (2`)
Thcn cquation 22 rcay mcan- that X
t
X
t
() i- a -tocha--
tic jrocc-- -ati-tyin cquation 2`
`ow wc wi conccntratc on thc cxi-tcncc ot

_
t
0
f(s, )dB
s
()

whcrc B
t
() i- 1-dimcn-iona Lrownian motion -tartin at
thc oriin
13
`ow -ujjo-c that 0 a < b and f(t, ) i- ivcn \c want
to dcnc
_
b
a
f(t, )dB
t
().
\c want to intcratc thc tunction f(t, ) B
t
() in thc
intcrva |0, b| Lcnc a -tcj tunction
1
(t, ) y

1
(t, ) B
t
i1
() t |t
i1
, t
i
| 1 i n
ic
1
(t, ) i- a -tcj tunction a- a tunction ot t and
1
(t, )
takc- thc vauc ot f in thc ctt cndjoint ot thc intcrva
|t
i1
, t
i
|
Lcnc anothcr -tcj tunction
2
(t, ) y

2
(t, ) B
t
i
() t |t
i1
, t
i
| 1 i n
ic
2
(t, ) i- a -tcj tunction a- a tunction ot t and
2
(t, )
takc- thc vauc ot f in thc riht cndjoint ot thc intcrva
|t
i1
, t
i
|
14
Ior thc tunction
1
wc ct
E|
_
b
0

1
(t, )dB
t
()| E|

j0
B
j
B
j
|

j0
E|B
j
(B
j+1
B
j
)|

j0
(E|B
j
B
j+1
| E|B
j
B
j
|)

j0
(E|B
j
B
j+1
| E|B
j
B
j
|)

j0
(
2
t
j

2
t
j
) 0.
Lut u-in thc tunction
2

E|
_
b
0

2
(t, )dB
t
()| E|

j0
B
j+1
B
j
|

j0
E|B
j+1
(B
j+1
B
j
)|

j0
(E|B
j+1
B
j+1
| E|B
j+1
B
j
|)

j0
(E|B
j+1
B
j+1
| E|B
j+1
B
j
|

j0
(
2
t
j+1

2
t
j
)

j0
(t
j+1
t
j
)
2
b.
15
Irom thi- wc -cc that wc can no oncr choo-c an aritrary
joint t |t
j
, t
j+1
| Thc toowin two choicc- havc turncd
out to c thc mo-t u-ctu onc-
1 Choo-c t t
j
tor t |t
j
, t
j+1
| (ic thc ctt cndjoint)
Thi- dcnc- thc Ito integral
2 Choo-c t (t
j
+ t
j+1
)/2 tor t |t
j
, t
j+1
| (ic thc mid-
joint ot thc intcrva) Thi- dcnc- Stratonovich integral
Thc lto intcra i- thc onc mo-t widcy u-cd (thouh in -omc
ca-c- thc Stratonovich intcra i- morc convcnicnt)
16
Irom now onc wc wi conccntratc on thc lto intcra `ow
wc mu-t nd out thc -ct ot tunction- wcrc thc ajjroximation
jroccdurc wi work out -uccc--tuy lt turn- out that it
work- wc it f ha- thc jrojcrty that cach ot thc tunction-
f(t
j
, ) only depends on the behaviour of B
s
() up
to time t
j
Thi- cad- to toowin conccjt-
Denition 2.1 Let B
t
() be n-dimensional Brownian
motion. Then we dene F
t
F
(n)
t
to be the -algebra
generated by the random variables B
s
() s t. In other
words, F
t
is the smallest -algebra containing all sets of
the form
{ B
t
1
() F
1
, , B
t
k
() F
k
}
where k 1, 2, . . . , t
j
t and F
t
j
R
n
are Borel sets.
Onc ottcn think- ot F
t
a- thc hi-tory ot B
s
uj to timc t
lt a tunction h R
n
i- F
t
mca-urac it mcan- that thc
vauc ot h() can c dccidcd trom thc vauc- ot B
s
() tor
s t
Ior cxamjc h() B
t/2
() i- F
t
mca-urac, whic h()
B
2t
() i- not Iinay, notc that F
s
F
t
tor s < t, ic {F
t
}
i- increasing
17
Denition 2.2 Let {N
t
}
t0
be an increasing family of
-algebras of subsets of . A process g(t, ) |0, |
R
n
is called N
t
-adapted if for each t 0 the function
h R
n
h() g(t, )
is N
t
measurable.
Thc jrocc-- g(t, ) B
t/2
() i- F
t
-adajtcd, whic thc jro-
cc-- g(t, ) B
2t
() i- not F
t
-adajtcd
18
Thc toowin dcnition, dcnc- thc ca-- ot tunction- tor
which thc lto intcra wi c dcncd
Denition 2.3 Let V V(a, b) be the class of functions
f(t, ) |0, ) R
n
such that:
1. (t, ) f(t, ) is B F measurable, where B de-
notes the Borel -algebra on |0, | and F denotes a
-algebra on .
2. f(t, ) is F
t
-adapted.
3. E|
_
b
a
f(t, )
2
dt| < .
19
2.1 The Ito integral
Thc idca ot dcnin thc intcra
I|f|()
_
b
a
f(t, )dB
t
()
whcrc B
t
i- 1-dimcn-iona Lrownian motion i- a- toow-
I|| i- dcncd tor -imjc ca-- ot tunction-
Thcn it i- -hown that cach f V can c ajjroximatcd
y -uch -
Iinay thi- i- u-cd to dcnc
_
fdB a- thc imit ot
_
dB
a- f (imit in L
2
(P))
20
A u-ctu ca-- ot tunction- i- dcncd a- toow-
Denition 2.4 A function V is called elementary
if it has the form
(t, )

j
e
j
()
[t
j
,t
j+1
[
(t)
where denote the characteristic (indicator) function.
`otc that thc tunction (t, ) i- a -tcj tunction a- a tunction
ot t -incc e
j
() i- ony a tunction ot in cach intcrva `otc
a-o that -incc V cach ot thc tunction- e
j
() mu-t c
F
t
j
-mca-urac
Ior thc tunction-
1
and
2
dcncd ctorc, thc tunction
1
i- F
t
j
mca-urcac whcrca-
2
i- not -incc
2
i- thc vauc ot
thc Lrownian motion at thc cnd ot thc intcrva |t
j
, t
j+1
|
Ior ccmcntary tuntion- (t, ) thc intcra i- dcncd a-
toow-
_
b
a
(t, )dB
t
()

j0
e
j
()|B
t
j+1
B
t
j
|(). (2o)
`ow it rcmain- to jrovc that cach tunction f V can
c ajjroximatcd y ccmcntary tunction- Ior thi- -cc |1!,
Ok-cnda|
21
Ior thc ccmcntary tunction wc havc
Theorem 2.1 (The Ito isometry) Let (t, ) be a bounded
elementary function then
E
_
__
b
a
(t, )dB
t
_
2
_
E
__
b
a

2
(t, )dt
_
.
Proof A- ctorc wc jut B
j
B
t
j+1
B
t
j
and dcnotc
B
t
j+1
a- B
j+1
Thcn
E|e
j
e
i
B
j
B
i
|
_
0 if i j
E|e
2
j
|(t
j+1
t
j
) if i j
-incc thc variac- e
j
e
i
B
j
and (B
i
) arc indcjcndcnt y
dcnition it i j Thu-
E
_
__
b
a
dB
_
2
_
E
__
b
a
dB
_
b
a
dB
_
(2)
E
_
_
_
_

j
e
j
B
j
_
_
_

i
e
i
B
i
_
_
_

i
E|e
j
e
i
B
j
B
i
| (2S)

i
E|e
2
i
t
j
| E
__
b
a

2
dt
_
.
22
Denition 2.5 (The Ito integral) Let f V(S, T ).
Then the lto intcra of f from S to T is dened by
_
T
S
f(t, )dB
t
() im
n
_
T
S

n
(t, )dB
t
() (limit in ) L
2
(P)
(29)
where {
n
} is a sequence of elementary functions such
that
E|
_
T
S
(f(t, )
n
(t, ))
2
dt| 0 as n (210)
Such a -cqucncc -ati-tyin thc aovc imit cxi-t- Thc vauc
ot thc intcra doc- not dcjcndt on thc actua choicc ot a-
on a- (210 hod- Iurthcrmorc, thc toowin imjortant
rc-ut toow-
Theorem 2.2 (The Ito isometry)
E
_
__
b
a
f(t, )dB
t
_
2
_
E
__
b
a
f
2
(t, )dt
_
f V(a, b)
.
Thcrc cxi-t- a tunction ot thc tyjc which convcrc- to f
unitormy, and it toow- thcn that thcorcm 22 i- a-o truc
23
Thc lto intcra can c cacuatcd u-in thc dcnition di-
rccty, that i-
Iind a -cqucncc {
n
} ot ccmcntary tunction- that aj-
jroximatc- f, ct (210)
-c (29) to nd thc intcra ot thc imit in L
2
(P) ot thc
intcra invovin
24
Thc toowin cxamjc iu-tratc- thc jrincijc
Example 2.1 Assume B
0
0. Then
_
t
0
B
s
dB
s

1
2
B
2
t

1
2
t (211)
Proof
Put
n
(s, )

j
B
j
()
[t
j
,t
j+1
)
(t), where B
j
B
t
j
.
Then
E|
_
t
0
(
n
B
s
)
2
ds| E|

j
_
t
j+1
t
j
(B
j
B
s
)
2
ds|

j
_
t
j+1
t
j
(s t
j
)ds

j
1
2
(t
j+1
t
j
)
2
0 as t
j
0
According to (2.9) the integral can be found as
_
t
0
B
s
dB
s
im
t
j
0
_
t
0

n
dB
s
im
t
j
0

j
B
j
B
j
25
Since
(B
2
j
) B
2
j+1
B
2
j
(B
j+1
B
j
)
2
+ 2B
j
(B
j+1
B
j
) (212)
(B
j
)
2
+ 2B
j
B
j
we have that
B
2
t

j
(B
2
j
)

j
(B
j
)
2
+ 2

B
j
B
j
Therefore

j
B
j
B
j

1
2
B
2
t

1
2

j
(B
j
)
2
Since

j
(B
j
)
2
t in L
2
as t
j
0, the result fol-
lows.
Note the extra term
1
2
t compared to ordinary integra-
tion.
26
2.1.1 Some properties of the Ito integral
Theorem 2.3 (Properties of the Ito integral) Let f, g
V(0, T) and let 0 a < b. Then
(i)
_
b
a
fdB
t

_
u
a
fdB
t
+
_
b
u
fdB
t
tor a.a.,
(ii)
_
b
a
(cf + g)dB
t
c
_
b
a
fdB
t
+
_
b
a
gdB
t
c constant tor a.a.,
(iii) E
__
b
a
fdB
t
_
0,
(iv)
_
b
a
f
2
dB
t
is F
b
measurable.
Thi- i- ca-y to jrovc tor ccmcntary tunction-, -o y takin
imit- wc otain thc rc-ut tor a f, g V(0, T)
27
An imjortant jrojcrty i- that the Ito integral is a mar-
tingale \c -ha rcturn to that joint attcr thc toowin
dcnition
Denition 2.6 Let (, F, P) be a probability space. A
ltration on (, F) is a family {M
t
}
t0
of -algebras
M
t
F such that
0 s < t M
s
M
t
(i.e. {M
t
} is increasing). An n-dimensional stochastic
process {M}
t0
on (, F, P) is called a martingale with
respect to a ltration {M}
t0
if
1. M
t
is M
t
-measurable for all t,
2. E||M
t
|| < ,
3. E|M
s
|M
t
| M
t
for all s t.
28
Example 2.2 Brownian motion B
t
in R is a martingale
with respect to the -algebra F
t
generated by {B
s
. s t}.
Clearly B
t
is F
t
-measurable and E||B
t
|| < since
E||B
t
||
2
E||B
t
|
2
| E|B
2
t
| B
2
0
+
2
t <
where the rst inequality follows from Jensens inequal-
ity. Now,
E|B
t+s
|F
t
| E|B
t+s
B
t
+ B
t
|F
t
|
E|B
t+s
B
t
|F
t
| + E|B
t
|F
t
| 0 + B
t
B
t
since B
t+s
B
t
is independent of F
t
and E|B
t+s
B
t
| 0.
29
Theorem 2.4 (The Ito integral is a martingale) If
f L
2
, the stochastic integral
X(t)
_
t
0
f(s, )dB
s
() (213)
is a martingale with respect to the ltration {M
t
}
t0
.
Proof Iir-t, notc that E|X(t)| < toow- trom X(t)
L
2
Ior thc ccmcntary tunction- it i- rcadiy -ccn that
E|X(t + s) X(t)|F
t
| E|X(t + s) X(t)| 0
Thc rc-ut toow- thcn y oin to thc imit -uch that thc
-tcjtunction- f
30
Theorem 2.5 Let f V(0, T). Then there exists a
t-continuous version of
_
t
0
f(s, )dB
s
() 0 t T
i.e. there exist a t-continuous stochastic process J
t
on
(, F, P) such that
P|J
t

_
t
0
fdB| 1 0 t T.
Irom now on it wc c a--umc that
_
t
0
f(s, )dB
s
() mcan-
a t-continuou- vcr-ion ot thc intcra
31
Theorem 2.6 Let f(t, ) V(0, T) for all T. then
X
t
X
0
+
_
t
0
f(s, )dB
s
() (21!)
is a martingale with respect to F
t
.
32
Thc tact that an lto intcra i- a martinac i- onc ot thc
rca-on- why lto intcra- arc thc mo-t widcy u-cd Actuay
thc rcvcr-c i- a-o truc Any F
t
martinac (with rc-jcct to
P
0
. whcrc P
0
thc jroaiity aw ot B
t
-tartin at 0) can
c rcjrc-cntcd a- an lto intcra Thi- rc-ut i- cacd thc
martingale representation theorem and i- a- toow-
Theorem 2.7 (Martingale representation theorem)
Suppose that M
t
is an F
t
martingale and that M
t

L
2
(P
0
) for all t 0. Then there exists a unique stochas-
tic process g(s, ) such that g V|0, t| for all t 0 and
M
t
() E|M
0
| +
_
t
0
g(s, )dB
s
() t 0.
Remark. L
2
-tand- tor a mca-urac tunction- with an
intcrac -quarc
33
2.1.2 Extensions of the Ito integral
2.1.3 Multi-dimensional Ito integral
Thi- rcquirc- a rcaxation ot thc mca-uraiity condition in
thc dcnition 23 \c -imjy con-idcr a arcr (and incrca--
in) tamiy ot -acra- H
t
-uch that
1 B
t
i- a martinac with rc-jcct to H
t
, and
2 f
t
i- H
t
-adajtcd
34
Lct B (B
1
, B
2
, . . . , B
n
) c n-dimcn-iona Lrownian mo-
tion and ct |
i,j
(t, )| c an mn matrix whcrc cach
cntry
i,j
(t, ) -ati-c- 1 and 3 in dcnition 23 and i- F
(n)
-
adajtcd Thcn u-in matrix notation wc dcnc
_
t
0
(t, )dB(t, )
_
t
0
_
_

11
(t, )
1n
(t, )

m1
(t, )
mn
(t, )
_
_
_
_
dB
1
(t, )

dB
n
(t, )
_
_
to c m1 matrix (ic coumn vcctor) who-c ith comjo-
ncnt i- a -um ot 1-dimcn-iona lto intcra-
n

j=1
_
t
0

i,j
(s, )dB
j
(s, ).
35
2.1.4 Ito integral dened by convergence in probability
Thi- rcquirc- a wcakcnin ot thc condition (iii) ot Lcnition
?? to
iii P|
_
t
0
f(s, )
2
ds < tor a t 0| 1 (21`)
Lct thc modicd ca-- ot tunction- c dcnotcd y W Thc
con-truction i- -ihty modicd
I|| i- dcncd tor -imjc ca-- ot tunction-
Thcn it i- -hown that cach f W can c ajjroximatcd
y -uch -
Iinay thi- i- u-cd to dcnc
_
fdB a- thc imit ot
_
dB
a- f (imit in jroaiity)
Thu- wc may dcnc
_
T
S
f(t, )dB
t
() im
n
_
T
S

n
(t, )dB
t
() (imit in) L
2
(P)
(21o)
A- ctorc a t-continuou- vcr-ion ot thi- intcra cxi-t-
36
2.1.5 A comparison of Ito and Stratonovich integrals
lt ha- ccn arucd that thc matcmatica intcrjrctation ot
thc cquation
dX
dt
b(t, X
t
) + (t, X
t
)W
t
. (21)
i- that X
t
i- a -oution ot thc intcra cquation
X
k
X
0
+
_
t
0
b(s, X
s
)ds +

_
t
0
(s, X
s
)dB

s
. (21S)
A- indicatcd jrcviou-y, thc lto intcra i- ,u-t onc intcrjrc-
tation ot

_
t
0
f(s, )dB
s
()

37
Thc Stratonovich intcra i- anothcr intcrjrctation, cadin
(in cncra) to dicrcnt rc-ut-
ln a comjari-on with thc ordinary (dctcrmini-tic) -oution
thc Stratonovich ha- an advantac thi- i- indicatcd in thc
toowin
Ior cach ct X
(n)
t
() c thc -oution ot thc corrc-jondin
(dctcrmini-tic) dicrcntia cquation
dX
dt
b(t, X
t
) + (t, X
t
)
dB
(n)
t
dt
(219)
Thcn X
(n)
t
() convcrc- to -omc tunction X
t
(), and it
turn- out that thi- -oution coincidc- with thc -oution o-
taincd y the Stratonovich integral
X
k
X
0
+
_
t
0
b(s, X
s
)ds +
_
t
0
(s, X
s
) dB
s
. (220)
38
Thi- imjic- that X
t
i- thc -oution ot thc toowin modied
Ito equation
X
k
X
0
+
_
t
0
b(s, X
s
)ds+
1
2
_
0
t

(s, X
s
)(s, X
s
)ds+
_
t
0
(s, X
s
)dB
s
.
(221)
whcrc

dcnotc- thc dcrivativc ot (x, t) with rc-jcct to


x Thc cquation- -how- that thcrc i- an cxjicit conncction
ctwccn thc two modc-
`oticc that thc two intcra- coincidc it (x, t) doc- not
dcjcnd on x
39
Thc toowin cxamjc iu-tratc- that thc Stratonovich intc-
ra corrc-jond- to thc ordinary (dctcrmini-tic) intcration
Example 2.3 Consider the stochastic integral:
_
t
0
B(s)dB(s) (222)
Dene
I
0
im

B(t
i
)(B(t
i+1
) B(t
i
)) (223)
I
1
im

B(t
i+1
)(B(t
i+1
) B(t
i
)) (22!)
Then, as we have shown previously,
I
1
I
0
im

(B(t
i+1
) B(t
i
))
2
t. (22`)
Hence, the result of the integral depends on the evalua-
tion point of the integrand.
Let us now consider
I

im

((1 )B(t
i
) + B(t
i+1
))B(t
i
) (22o)
im

_
B(t
i
)B(t
i
) + (B(t
i
))
2

(22)
im

_
1
2
(B
2
(t
i+1
) B
2
(t
i
)) + (
1
2
)(B(t
i
))
2
_
(22S)

1
2
(B
2
(t) B
2
(0)) + (
1
2
)t (229)
Now it is clearly seen that for 1/2, which correponds
to the Stratonovich intcra, the same result as for the
ordinary (deterministic) case is obtained.
40
`LXT
2.1.6 The Ito process and Ito formula
Thc a-ic dcnition ot lto intcra- i- not vcry u-ctu whcn
wc try to cvauatc a ivcn intcra Thi- i- -imiar to thc
-ituation tor ordinary licmann intcra-, whcrc wc do not
u-c thc a-ic dcnition, ut rathcr thc tundamcnta thcorcm
ot cacuu- ju- thc chain ruc in thc cxjicit cacuation- lt
turn- out that i- jo--ic to c-tai-h an Ito integral version
of the chain rule, called Ito formula, and it i- vcry u-ctu
tor cvauation ot lto intcra-
41
Bibliography
|1| A-jund L and Lunart L, A rst course in Integra-
tion Lot, linchart and \in-ton lnc nitcd Statcd ot
Amcrica 19oo
|2| L,orn-on O!, Fyrirlestrar Lkindarfraedi, fyrri par-
tur ln-titutc ot mathcmatic-, Thc nivcr-ity ot lccand,
lcyk,avik 1990
|3| Lrnd-tcd A, Optimering og Konveksitet `athc-
matica ln-titutc, Cojcnhacn nivcr-ity, Cojcnhacn
1990
|!| Lurki !C, A rst Course in Mathematical Analysis
Camridc nivcr-ity lrc-- 19SS
|`| Chri-tcn-cn `, Obligations investering 1st edition
!uri-t o Okonomtorundct- Iora 1990
|o| Cox !C, lncr-o !L and lo-- SA, A lc-
cxamination ot Traditiona Lyjothc-c- aout thc Tcrm
Structurc ot lntcrc-t latc- Journal of Finance, vo
XXX\l, `o !, Scjtcmcr 19S1 j o9
|| Conrad-cn I, En introducion til statistik bind 1A
and 1B, 6th edition ln-titutc tor mathcmatica mod-
42
cin, Thc Tcchnica nivcr-ity ot Lcnmark, Lyny
199`
|S| Ldward- CL and lcnncy LL, Calculus and ana-
lytic geometry, third edition lrcnticc-La lntcrna-
tiona Ldition-, `cw !cr-cy 1990
|9| Grimmct Gl and Stirzakcr Ll Probability and
Random Processes, Oxtord nivcr-ity lrc-- 19S9
|10| !rcn-cn lL, American Option Pricing Lcjartmcnt
ot Iinancc, Thc Aarhu- Schoo ot Lu-inc--, Aarhu-
199!
|11| `ichc-cn A Investering og konomisk levetid, Lc-
jartmcnt ot lndu-tria `anacmcnt and Lninccrin ,
Thc Tcchnica nivcr-ity ot Lcnmark, Lyny 199`
|12| lavn L, Noter til statisk og dynamisk optimerin, ln-
-titutc ot mathcmatica modcin, Thc Tcchnica ni-
vcr-ity ot Lcnmark, Lyny 199o
|13| Stctan--on !l Um dirun a fallinu x
_
h
(x)
g
(x)f(x, t)dt `athcmatic ln-titutc, Thc nivcr-
-ity ot lccand, lcyk,avik, 1991
|1!| Ok-cnda L, Stochastic Dierential Equations,
fourth edition Sjrincr-\cra, Lcrin Lcidccr
199`
43

You might also like