Scribd
Upload
289 views1 page

Kolmogorov Forward

The document derives the Kolmogorov forward equation from the Kolmogorov backward equation. It shows that if the transition density p satisfies the Kolmogorov backward equation in the initial variables x and s, then p also satisfies the dual equation in the terminal variables y and t, known as the Kolmogorov forward equation. This establishes that the transition density p satisfies both the Kolmogorov backward and forward equations.

Uploaded by

suedesuede
Copyright
© © All Rights Reserved
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
289 views1 page

Kolmogorov Forward

The document derives the Kolmogorov forward equation from the Kolmogorov backward equation. It shows that if the transition density p satisfies the Kolmogorov backward equation in the initial variables x and s, then p also satisfies the dual equation in the terminal variables y and t, known as the Kolmogorov forward equation. This establishes that the transition density p satisfies both the Kolmogorov backward and forward equations.

Uploaded by

suedesuede
Copyright
© © All Rights Reserved
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/ 1

Kolmogorov Forward Equation.

Also by the Markov property, it immediately follows that


Z
(x,s)
u(x, s) = E f (Xt ) = p(x, s; y, t)u(y, t) dy
We derived the Kolmogorov backward equation in class. This short note deduces
the Kolmogorov forward equation from the Kolmogorov backward equation. Let X
be a diffusion satisfying the SDE for any t ∈ [s, T ]. (Alternately, you can deduce the above equality using uniqueness
of solutions to (1).) Now, differentiating both sides in t, and using (1) gives
dXt = b(Xt ) dt + σ(Xt ) dWt , Z
0 = ∂t u(x, s) = [∂t p(x, s; y, t)u(y, t) + p(x, s; y, t)∂t u(y, t)] dy
where b and σ are time independent and Lipshitz. Suppose further X has a (smooth) Z
transition density
(x,s) = [∂t p(x, s; y, t)u(y, t) − p(x, s; y, t)Ly u(y, t)] dy
p(x, s; y, t) = P (Xt ∈ dy).
Z
∂t p(x, s; y, t) − L∗y p(x, s; y, t) u(y, t) dy
 
We know that p satisfies the Kolmogorov backward equation in the initial variables =
x and s. Namely,
Note, to obtain the last inequality we had to integrate by parts. All the boundary
∂s p + Lx p = 0, and lim p(·, s; y, t) = δy ,
s→t− terms involved are 0 because one can show that p vanishes at infinity.
Now choosing t = T , we see that
where X 1X Z
L= bi ∂i + ai,j ∂i,j
∂t p(x, s; y, T ) − L∗y p(x, s; y, T ) f (y) dy = 0.
 
i
2 i,j

and X Since f ∈ C02 is arbitrary, we are done.


ai,j = ai,j (x) = σi,k (x)σj,k (x).
k

The Kolmogorov forward equation says that p satisfies the dual equation in the
variables y, t.
Proposition 1 (Kolmogorov forward equation). Let L∗ be the dual of L, defined by
X 1X
L∗ g = −∂i (bi g) + ∂i,j (ai,j g).
i
2 i,j

Then
∂t p + Ly p = 0, and lim p(x, s; ·, t) = δx ,
t→s+

Remark. The operator L∗ is the dual of L with respect to the L2 inner product.
Namely, if f, g ∈ Cc2 (Rd ), then
Z Z
Lf (x) g(x) dx = f (x) L∗ g(x) dx
Rd Rd

Proof. Fix T > 0, and always assume 0 6 s < t 6 T . Let f ∈ Cb2 (Rd ), and define
Z
(x,s)
u(x, s) = E f (XT ) = p(x, s; y, T )f (y) dy.

We know that u satisfies the Kolmogorov backward equation

∂s u + Lx u = 0 and u(x, T ) = f (x). (1)

You might also like