Sei sulla pagina 1di 9

Annals of Studies in Science and Humanities

Volume 1 Number 1 (2015) : 19


http://journal.carsu.edu.ph/
Online ISSN 02408-3631

A Characterization of Backwards It Integral to a Set of


Singularities in

Jayrold P. Arcede1 and Emmanuel A. Cabral2


1
Department of Mathematics, Caraga State University, Butuan City, Philippines
2
Department of Mathematics, Ateneo de Manila University, Quezon City, Philippines

Received: September 15, 2014 Accepted: September 23, 2015

ABSTRACT

In this paper, we give a characterization between backwards It integrability and the set of
singularities in .

Key words: AC 2 -property, backwards It integral, backwards L2 -martingale

1 Introduction
Arcede and Cabral (2011) introduced a stochastic integral called the backwards It
integral which assumed adaptedness property with respect to backwards filtration. This
integral was defined using the generalized Riemann approach due to Henstock and Kurzweil
in 1950s. It was also called the Henstock approach, a technique applied in the classical
non-stochastic integral, see Henstock (1955, 1988) and Kurzweil (1957).
Henstock approach diers from the usual Riemann as it uses non-uniform meshes, that
is, the sizes of the intervals vary from point to point. This minor modification leads to a
more general integrals compared to that of Riemann-Stieltjes integral or even the Lebesgue-
Stieltjes integral (Henstock, 1955, 1988). It is known that a stochastic integrals cannot be
defined using the Riemann approach for integrators can have paths of unbounded variation.
Moreover, integrands can be highly oscillatory (Arcede, 2011).
However, It integral can be defined using the generalized Riemann approach, see for
instance Stojanovic (1972), McShane (1974), Henstock (1991), and Lee (1993). It is known

Corresponding Author
Email:jparcede@carsu.edu.ph
J.P. Arcede & E.A. Cabral Vol. 1 No. 1 2015

also that the resulting integrals encompass the classical one, see Toh and Chew (2001,
2003). In this note, we shall characterize backwards It integrability with regards to ,
whose collection of tags in its interval point pair can be considered as a set of singularities.
This work follows closely that of Cabral (1999).

2 Preliminaries
We will assume familiarity with the definitions and basic properties including the Hen-
stock Lemma of the backwards It integral. Moreover, the setting is the same as in authors
paper in Arcede and Cabral (2011). Throughout this note, R denotes the set of real num-
bers, R+
0 the set of nonnegative real numbers, N the set of positive integers and (, G, P)
denotes a probability space.
Definition 2.1. Let {G s : 0 s T } be a family of sub -algebras of G. {G s : 0 s T }
is called a backwards filtration if G t G s for all 0 s < t T . If in addition,
{G s : 0 s T } satisfies the following condition:
(1) G T contains all sets of P -measure zero in G; and

(2) for each s [0, T ], G s = G s where G s = >0 G
s .

Then {G s : 0 s T } is called a standard backwards filtration.


We often write {G s } instead of {G s : 0 s T }. Some terminologies above can be
found in Applebaum (2004).
Definition 2.2. A stochastic process f or simply process is a function f : [0, T ] R,
where [0, T ] is an interval in R+0 and f (, s) is G -measurable for each s [0, T ]. A process
s

f = {fs : s [0, T ]} is said to be adapted to the standard backwards filtration {G s } if fs is


G s -measurable for each s [0, T ].
{ }
Let B = Bt : t R+ 0 be a standard Brownian motion (BM). Let (Bu : s u T )
be the smallest -algebra generated by {Bu : s u T }. This is the smallest -algebra
containing the information about the structure of BM on [s, T ].
Throughout this note, we assume that the standard backwards filtration {G s } is the
family of -algebras (Bu : s u T ). This family is then called the natural backwards
filtration of B. Let (, G, {G s }, P) be a standard backwards filtering space. We write Lp ()
for Lp (, G, P) where f Lp () if E(f )p < . For f L1 (), let E(f ) denote the
expectation of f , that is, E(f ) = f dP. The conditional expectation of f given G s is the
random variable E(f |G s ).

3 Backwards It Integral
In this section, we shall present the backwards It integral and some related results
needed to prove the main result.

2
Annals of Studies in Science and Humanities Vol. 1 No. 1 2015

Definition 3.1. Let be a positive function on (0, T ]. A finite collection D = {((ui , i ] , i )}ni=1
of interval-point pairs is said to be a backwards partial division of [0, T ] if {(ui , i ]}ni=1 is a
finite collection of disjoint subintervals of (0, T ].

Definition 3.2. An interval-point pair ((u, ], ) is said to be backwards -fine if (u, ]


( (), ], whenever (u, ] (0, T ] and (0, T ]. We call D = {((ui , i ], i )}ni=1 a
backwards -fine partial division of [0, T ] if D is a backwards partial division of [0, T ] and
for each i, the interval-point pair ((ui , i ], i ) is backwards -fine.

Let > 0. One may not be able to find a full division that covers the entire interval
(0, T ]. For example, take () = 2 . Then the interval (0, T ] cannot be covered by any
f inite collection of backwards -fine intervals.

Definition 3.3. Let > 0 be given, a backwards -fine partial division D is said to fail to
cover (0, T ] by at most Lebesgue measure if


n

T (i ui ) .
i=1

The backwards It integral is define as follows.

Definition 3.4. Let f = {fs : s [0, T ]} be a process adapted to the standard backwards
filtering space (, G, {G s }, P). Then f is said to be backwards It integrable on [0, T ] if there
exists an A L2 () such that for any > 0, there exist a positive function on (0, T ] and a
positive number such that for any backwards -fine partial division D = {((ui , i ], i )}ni=1
of [0, T ] that fails to cover (0, T ] by at most Lebesgue measure we have
( )
E |S(f, D, , ) A|2 , (1)

T
n
where S(f, D, , ) = i=1 fi (Bi Bui ). We denote A by ft dBt .
0

The following definition can be found in Tay et al. (2001).

Definition 3.5. Let F = {Fs : s [0, T ]} be a stochastic process. Then the process
F( is said to have )an AC 2 - property if for each > 0, there exists > 0 such that
2
E ni=1 F (ui , vi ) , for any finite collection of disjoint subintervals {(ui , vi ]}ni=1 of

[0, T ] for which ni=1 |vi ui | .
T
Theorem 3.1. Let f be backwards It integrable on [0, T ]. Let (, T ) = ft dBt . Then
has AC 2 -property.

3
J.P. Arcede & E.A. Cabral Vol. 1 No. 1 2015

4 A Characterization using
In order to dierentiate a function with respect to another function, Henstock intro-
duced in Henstock (1963) a concept called inner variation. Roughly speaking, variation is
defined over all interval-point pairs whereas inner variation is defined over certain family
of interval-point pairs (Lee, 2007). The following theorem is due to Cabral (1999):

Theorem 4.1. A function is Henstock integrable on [a, b] with primitive F if and only if
for every > 0 there is (x) > 0 for x [a, b] such that for any -fine partial division D
in we have

(D) |F (v) F (u)| ,

and

(D) |f (x)(v u)| (2)

where
= {([u, v], x) : |F (v) F (u) f (x)(v u)| |v u|} (3)

The condition above is known as the double Lusin condition. It is an extension of


inner variation. The idea of came from Lu JiTan (Lu and Lee, 1999), independently of
Henstock.
The aim of this section is to characterize a backwards It-integrable process f on [0, T ]
with primitive F in relation to a set X on [0, T ] defined in (4) below, and of the set of
interval-point pairs . To be concise, given any > 0 there exists a positive function on
[0, T ] and a positive number such that the interval [0, T ] is divided into [0, T ]\X. One can
considered the set X as a set of singularities in some sense on [0, T ]. The reader is advised
to see Lu and Lee (1999), where the set of singularities is fixed for all > 0 and Cabral
(1999) where the concept is carried on n-dimensional space but the set of singularities is
not fixed. Here, we apply the concept with regards to backwards It-integrability.

Definition 4.1. Let f and F be adapted processes on [0, T ]. For any given > 0 and a
positive function on [0,T] , let be the set of interval-point pairs ((u, ], ) such that
)

E (| F (u, ) f ()(B Bu ) 2 E(B Bu )2 = ( u).

Now, define the set X to be a subset of [0, T ] such that for every point X there exist
an interval-point pairs ((u, ], ) which is backwards -fine. In set-theoretic notation,

X(, ) = { [0, T ] : there is a backwards -fine ((u, ], ) } . (4)

We remark that the collection of all backwards -fine interval-point pairs ((u, ], )
forms a cover for X(, ). In brief, we write X for X(, ).

4
Annals of Studies in Science and Humanities Vol. 1 No. 1 2015


Theorem 4.2. Let f and F be adapted processes on [0, T ], where F (u, v) = uv f dBt .
Then f is backwards It-integrable on [0, T ] to F if and only if F satisfies AC 2 -property
and for every > 0 there exists a positive function on [0, T ] such that whenever D =

{(u, ], )} is a backwards -fine partial division of we have E (|(D) F (u, )|)2 and

E (|(D) f ()(B Bu )|)2
Proof. (=) Suppose F satisfies AC 2 -property and the two inequalities are true. We will
show that f is backwards It-integrable with primitive F on [0, T ]. Let D = {(u, ], )}
be a backwards -fine partial division of [0, T ] that fails to cover [0, T ] by at most a set of
measure . Then
( )2

E (D) F (u, ) f ()(B Bu ) 2(D \ ) E (|F (u, ) f ()(B Bu )|)2 +
( )2

4E (D ) F (u, ) +
( )2

4E (D ) f ()(B Bu )

2 ( u) + 4 + 4
2(T 0) + 8
= 10(T + 1).
Now note that since D fails to cover [0, T ] by at most a set of measure . Then the part
of [0, T ] not covered by D is of measure at most . Let this part be a finite collection of
N
disjoint subintervals {(si , ti ]}N
i=1 and i=1 (ti si ) < . Since F satisfies AC -property,
2
( N )
2

E F (si , ti ) .
i=1
Hence,
( 2 )

E (D) f (B Bu ) F (0, T )
( ( N )
2 ) 2

2E (D) f (B Bu ) F (u, ) + 2E F (si , ti )
i=1
2(T + 1) + 2
= 2(T + 2).
(=) Let f be backwards It-integrable on [0, T ] with primitive F and let
Ek = { [0, T ] : k 1 |f ()| < k} .
Let > 0. Then for each k N, there exist a positive function k () on [0, T ] and a positive
number k , such that
( )2 2
E (Dk ) |F (u, ) f ()(B Bu )|
k 2 2k+1

5
J.P. Arcede & E.A. Cabral Vol. 1 No. 1 2015

whenever Dk is a backwards k -fine partial division of [0, T ] that fails to cover [0, T ] by at
most a set of measure k . A positive function () may be chosen such that

() k () if Ek

and also the positive number in such a way that k . Therefore, for each backwards
-fine partial division D of [0, T ], we have
( )2

E (D) f ()(B Bu ) = (D) E (|f ()(B Bu )|)2
k2
(D) E (|(B Bu )|)2
E k
2


k2
< E (D) |F (u, ) f ()(B Bu )|
k=1
E k

k2 2

k=1
k 2 2k+1

= .
2
Furthermore, may be appropriately chosen so that
( )2 ( )2

E (D) F (u, ) 2E (D) F (u, ) f ()(B Bu ) +
( )2

2E (D) f ()(B Bu )

+ = .
2 2
The first term of the inequality above is due to Henstock Lemma. The proof is now
complete.

REMARK 1. In Theorem 4.2, the inequality involving the function f may be restated as
one which does not involve f. This restatement gives rise to Theorem ?? which is actually
equivalent to Theorem 4.2. The equivalence is proved by showing that the inequality
( )2

E (D) (B Bu )

leads to ( )2

(E (D) f ()(B Bu )

which is precisely the inequality appearing in Theorem 4.2.

6
Annals of Studies in Science and Humanities Vol. 1 No. 1 2015


Theorem 4.3. Let f and F be adapted processes on [0, T ], where F (u, v) = uv f dBt . Then
f is backwards It-integrable on [0, T ] to F if and only if F satisfies AC 2 -property and for
every > 0 there exists a positive function on [0, T ] such that whenever D = {(u, ], )}
is a backwards -fine partial division of we have

E (|(D) F (u, )|)2 and E (|(D) (B Bu )|)2

Proof. (=) First, observe the following:



( )2

E (D) (B Bu ) = (D)E |(Bk Buk )|2 + 2 (Bk Buk )(Bl Bul )
k k<l

= (D) E(Bk Buk )2 + 2 E {(Bk Buk )(Bl Bul )}
k k<l

= (D) E(Bk Buk )2
k

Note that the second term in the equation above is zero due to the orthogonality of incre-
ments of BM.
Now, let f be backwards It-integrable to F on [0, T ]. Then for each > 0, there exist
a positive function () on [0, T ] and a positive number , such that
( )2
E (D) |F (u, ) f ()(B Bu )| 2

whenever D is a backwards -fine partial division of [0, T ] that fails to cover [0, T ] by at
most a set of measure . Therefore, for each backwards -fine partial division D of
[0, T ], we have

( )2
1
E (D) (B Bu ) = (D) E(B Bu )2

)
1
(D) E (| F (u, ) f ()(B Bu ) 2
(5)
1 2


= .

Therefore,
(D) ( u) < . (6)

By Theorem 3.1, F satisfies AC 2 -property. Hence, choosing = in Definition 3.5 makes


the following straightforward:
( )2

E (D) F (u, ) ,

7
J.P. Arcede & E.A. Cabral Vol. 1 No. 1 2015

whenever {(u, ]} is a finite collection of disjoint subintervals of [0, T ] satisfying Eq.(??).


(=) Let
Ek = { [0, T ] : k 1 |f ()| < k}
as defined previously.
This part of the proof shows that the two inequalities on f and F in Theorem 4.2 are
satisfied and f is backwards It integrable with primitive F.
Let > 0. Then for every k > 0, k N, there exist a positive function k () on [0, T ]
such that ( )2

E (Dk ) F (u, ) k

and ( )2

E (Dk ) (B Bu ) k

whenever Dk is a backwards k -fine partial division of [0, T ] in k , where k = k2 2k+1
and
{ ) }

k = ((u, ], ) such that E (| F (u, ) f ()(B Bu ) 2 k E(B Bu )2

A positive function () may be chosen such that () k () if Ek .



Therefore, for each backwards -fine partial division D k=1 k of [0, T ], we
have
( )2

E (D) f ()(B Bu ) = (D) E (|f ()(B Bu )|)2

k 2 (D) E (|(B Bu )|)2
Ek


< k2
k=1
k 2 2k+1

= .
2
The proof is now complete.

Acknowledgment
The authors would like to acknowledge the financial support from National Research
Council of the Philippines(NRCP) under Research Project No. B-110.

References
Applebaum, D. (2004). Lvy Processes and Stochastic Calculus. Cambridge University
Press.

8
Annals of Studies in Science and Humanities Vol. 1 No. 1 2015

Arcede, J. P. and Cabral, E. A. (2011). An Equivalent Definition for the Backwards


It Integral, Thai Journal of Mathematics 9 (3), 619-630.

Cabral, E. A. and Lee, P. Y. (2000/2001). A Fundamental Theorem of Calculus for


the Kurzweil-Henstock Integral in Rm . Real Anal. Exchange 26 (2), 867-876.

Chew, T. S., Tay J. Y. and Toh, T. L. (2001/2002). The non-uniform Riemann Ap-
proach to Its Integral. Real Anal. Exchange 27 (2), 495-514.

Henstock, R. (1963). Theory of Integration. Butterworths, London.

Henstock, R. (1988). Lectures on the Theory on integration. World Scientific, Singa-


pore.

Henstock, R. (1955). The eciency of convergence of factors for functions of a contin-


uous real variable. Journal London Math. Soc. 30, 273-286

Henstock, R. (1991). The General Theory of Integration. Oxford Science.

Henstock, R. (1990/1991). Stochastic and other functional integrals. Real Anal. Ex-
change 16, 460-470.

Kurzweil, J. (1957). Generalized ordinary dierential equations and continuous depen-


dence on a parameter. Czechoslovak Math. Journal 7, 418-446.

Lee, P.Y. (2007). Integral A` La Henstock. Scientiae Mathematicae Japonicae Online


763-777.

Lu, J. T. and Lee, P. Y. (1999). The Primitives of Henstock Integrable Functions in


Euclidean Space, Bull. London Math. Society 31, 173-180.

McShane, E. J. (1974). Stochastic Calculus and Stochastic Models. Academic Press,


New York.

Pop-Stojanovic, Z. R. (1972). On McShanes belated stochastic integral, SIAM J.


Appl. Math. 22, 89-92.

Toh, T. L. and Chew, T. S. (2003). The Riemann approach to stochastic integration


using non-uniform meshes, Journal of Mathematical Analysis and Application 280,
133-147.

Tay, J. Y., Toh, T. L., and Chew, T. S. (2001). The non-uniform riemann approach
to Its Integral, Real Analysis Exchange 27 (2), 495-514.

Xu, J. G. and Lee, P. Y. (1992). Stochastic integrals of It and Henstock, Real Anal.
Exchange 18, 352-366.

Potrebbero piacerti anche