Strong Taylor Schemes for Stochastic Volatility
Ito and Stratonovich Stochastic Calculus
Ito-Stratonovich drift conversion
Strong Numerical Schemes for SDE
Milstein scheme for commutative noise
Approximations of Volatility Models
General 2D Milstein scheme for stochastic volatility models
Approximations of the Double Integral
Subdivision (Kloeden - IC = 0)
Simulation of the Double Integral
Formulae derivation for Heston Volatility
Derivation of the 2D Milstein Scheme
We start with a formal definition of our process which we use the rest of the thesis. An N-dimensional Ito SDE with an M-dimensional Wiener process, e.g.
with M pair-wise independent scalar Wiener processes
as its components, will be written in vector form as:

or more general, as a component-wise:
(3)
This vector SDE could thus be written as:
dXt = A(Xt, t)dt + B(Xt, t)dWt
Similar notation can be used for a Stratonovich SDE:
dXt = A(Xt, t)dt + B(Xt, t) ◦ dWt
Ito Stochastic Calculus
For a sufficiently smooth transformation U : [0, T ] × RN --> R of the solution Xt of the Ito {SDE-3}:

f or i = 1, ..., N ; j = 1, ..., M
the scalar process Yt = U (Xt, t) satisfies the following Ito SDE:
(4)
with the differential operators L0, L1, :::, LM with respect to this SDE:

and

These operators play a fundamental role in Ito stochastic calculus through the Ito formulae for the stochastic chain rule and subsequently for stochastic Taylor expansions and numerical schemes for the SDEs that are based on stochastic Taylor expansions. It differs from what might be expected from deterministic calculus by the presence of the second order term in the L0 operator, which is essentially due to the fact that E((W )2) = Δt for the increment of a Wiener process over an interval of length Δt.
Stratonovich Stochastic Calculus
For a sufficiently smooth transformation U : [0, T ] × RN --> R of the solution Xt of the Stratonovich SDE:
(5)
f or i = 1, ..., N ; j = 1, ..., M
the scalar process Yt = U (Xt, t) satisfies the following Stratonovich SDE:
(6)
where the terms are all evaluated at (Xt, t). In operator form, this is:

The L0 operator of Ito calculus needs to be changed in Stratonovich calculus to:

while the Lj operators of Ito calculus remain unchanged in Stratonovich calculus.
Prof. Klaus Schmitz

Strong Taylor Schemes for Stochastic Volatility
This method requires formulas that are not always easy or possible to find. In this document, we present the corresponding approximations for both Euler and Milstein schemes for the usual Geometric Brownian Motion and the stochastic volatility models. Also, we present five methods of how we can simulate the double integrals for the 2 dimensional Milstein approximation.
By Prof. Klaus Erich Schmitz Abe