5
$\begingroup$

In Wikipedia the definition of generator of a semigroup is given as follows

For a Feller process $(X_t)_{t\ge0}$ with Feller Semigroup $T=(T_t)_{t\ge0}$ and state space $E$ we define the generator $(A,D(A))$ by: $$D(A)=\left\{f\in C_\infty :\lim_{t↓0}\frac{T_tf-f}{t}\text{exists as a uniform limit}\right\},$$

enter image description here

I saw another definition of infinitsimal generator in context of markov chains which says that if

$$d\phi(X_t)=A \phi(X_t)dt+dM_t$$

where $M_t$ is a martingale and $\phi \in C_c^{\infty}(\mathcal{X})$ and {$X_t$} is a markov chain on $\mathcal{X}$ then A is the infinitesimal generator. Are these two definitions equivalent?

$\endgroup$
2

1 Answer 1

2
$\begingroup$

I think this theorem is useful, it is from Continuous Time Markov Processes by Liggett:

Theorem: Let $X(t)$ be a Feller process on a state space $S$ and let $A$ be its generator with domain $D(A)$. Let $(P_x)_{x \in S}$ be the collection of path distributions - in case you are not familiar with this general definition, $P_x$ is the measure under which $X$ starts at $x$ almost surely. Then for every $x \in S$ and $f \in D(A)$: $$M(t) := f(X(t)) - \int_{0}^{t} (Af)(X(s)) ds$$ is a martingale under the measure $P_x$.

Note that the Hille-Yosida-theorem states that every Feller process corresponds to a generator and vice versa. However, not every Markov process is a Feller process - which in addition to the Markov property satisfies the following condition:

Feller Property: $\forall f \in \mathcal{C}(S)$ and $t\ge 0$ the function $\phi: S \rightarrow \mathbb{R}$ given by $$\phi(x) = \mathbb{E}_{P^{x}}[f(X(t))]$$ is also in $\mathcal{C}(S)$, i.e. continuous and vanishing at infinity.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.