Reversible Markov Chain
In the previous post, I introduced some of important Markov chain properties.
I would like to write about a Markov chain property that is used in Markov Chain Monte Carlo(mcmc), which is reversibility.
Reversibility
Let is a variable of a Markov chain. If there is a probability distribution such that:
Then the Markov chain is said to be reversible.
One of trivial examples of reversible Markov chain is as follows:
A reversible Markov chain has a stationary distribution
When defining as follows:
The equation of revesibility can be written as follows:
for all states of a Markov chain
Since the equation is true for all , when we sum up the left- and right-hand sides of the equation.
We get follows:
It means that we get the following equation.
This is the definition of a stationary distribution .