Take the stochastic process
$$
d x_t = \mu(t, x) dt + \sigma(t, x) d W_t
$$
where
The partial differential operator (infinitesimal generator) associated with the stochastic process is
\begin{align} \tilde{L_1} \equiv \tilde{\mu}(t, x) \partial_x + \frac{\tilde{\sigma}(t, x)^2}{2}\partial_{xx} \end{align}
Then, if the payoff in state
We can combine these to form the operator,
\begin{align}
\tilde{L} = \rho - \tilde{L_1}
\end{align}
and the boundary condition operator (using the
which leads to the PDE,
$$
\partial_t \tilde{u}(t,x) = \tilde{L}_t \tilde{u}(t,x) - \tilde{c}(t,x)
$$
and boundary conditions at every
As a numerical example, start with something like
$x^{\min} = 0.01$ $x^{\max} = 1.0$ -
$\tilde{\mu}(t,x) = -0.1 + t + .1 x$ - Note, that this keeps
$\tilde{\mu}(t,x) \geq 0$ for all$t,x$ . Hence, we know the correct upwind direction.
- Note, that this keeps
-
$\tilde{\sigma}(t,x) = \bar{\sigma} x$ for$\bar{\sigma} = 0.1$ $\tilde{c}(t,x) = e^x$ $\rho = 0.05$
Do a discretization of the
The stationary solution, at a
Given this solution, we can solve for the transition dynamics by going back in time from the