Information Technology Reference
In-Depth Information
Example 15.4.4
Consider the norm for the Bates model (9.24). Then,
w
(
1
,
0
)
1
(x
1
)
=
1,
w
(
1
,
0
)
2
(x
2
)
=
x
2
,
w
(
0
,
1
)
(x
1
)
=
w
(
0
,
1
)
1, as well as
w
(
0
,
0
)
1
(x
2
)
=
(x
1
)
=
1 and
1
1
2
w
(
0
,
0
)
2
x
2
.
(x
2
)
=
+
According to Proposition
15.4.2
and Corollary
15.4.3
, we define diagonal pre-
conditioners as follows, see also [19, Sect. 5]. Denote by
D
(i)
w
α
the diagonal matrix
(corresponding to the
i
th coordinate direction)
D
(i)
w
α
(
i
,k
i
),(
i
,k
i
)
:=
2
2
α
i
i
(w
i
)
2
(
2
−
i
k
i
)δ
i
,
i
δ
k
i
,k
i
∈ R
dim
V
L
×
dim
V
L
,
and set
∈ R
N
L
×
N
L
.
D
(
1
)
w
α
D
(d)
w
α
⊗···⊗
D
w
α
:=
(15.27)
|
α
|
≤
1
1
For the matrix
B
=
λ
M
+
k/
2
A
in (
15.23
), now define the preconditioner
:=
k/
2
D
w
α
1
/
2
.
D
(λ)
I
+
(15.28)
The next lemma is proven in [80].
Lemma 15.4.5
Let the assumptions of Corollary
15.4.3
hold
.
Assume the bilinear
form a(
·
,
·
)
:
V
×
V
→ R
satisfies
(3.8)
-
(3.9)
and assume
,
without loss of gener-
ality
,
that C
3
=
is equipped with the
norm given in
(
15.26
).
Let the matrices
B
and
D
be given by
(
15.23
)
and
(
15.28
),
respectively
.
Then
,
the preconditioned matrix
B
0
in
(3.9).
Further assume that the space
V
D
−
1
BD
−
1
,
:=
satisfies
:
There exists a constant c independent of L
,
λ and k such that
λ
min
(
B
+
B
H
)/
2
B
−
1
≥
c.
(15.29)
symmetric, the quantity
λ
min
(
B
+
B
H
)/
2
B
∈ R
N
L
×
N
L
−
1
is
equal to 1
/κ(
B
)
. Thus, estimate (
15.29
) is equivalent to the boundedness (from
above) of the condition number of
B
.
Note that for
A
Example 15.4.6
(CEV model) Consider the CEV model (4.17) with 0
<ρ<
0
.
5.
According to Proposition 4.5.1, the preconditioner for stiffness matrix
A
of this
model is given by
D
D
1
)
1
/
2
. Figure
15.1
shows the condition number of
:=
(
D
x
ρ
+
D
−
1
AD
−
1
for different values of
ρ
.
By combining Theorem 13.3.1 with Lemma
15.4.5
, we obtain the following con-
vergence result for the approximated option price in the Bates model.
Theorem 15.4.7
Let the assumptions of Theorem
13.3.1
and Lemma
15.4.5
hold
.
Then
,
choosing the number and order of time steps such that M
=
r
=
O
(L) and
Search WWH ::
Custom Search