Rylan Schaeffer

Logo
Resume
Research
Learning
Blog
Teaching
Jokes
Kernel Papers


Markov Process

Definition

A Markov process is a discrete-time stochastic process. To constuct an MP, 3 components are necessary:

  1. A set \(\mathcal{X}\), commonly called “states”
    2: An initial probability distribution over the set of states: \(p_0\) 3: A transition operator \(T: \mathcal{X} \rightarrow \mathbb{P}(\mathcal{X}}\), which specifies a conditional distribution
\[T(X_t = x) := P(X_{t+1} = x' \lvert X_t = x})\]

A Markov process is a set of random variables indexed by time \(\{X_t\}_{t=1, 2, 3...}\) where the distribution is given by:

\[\begin{align*} X_1 &\sim p_0\\ X_{t+1} \lvert X_{t} &\sim T X_t \end{align*}\]

If the set \(\mathcal{X}\) is finite, the transition operator can be written as a \(\lvert \mathcal{X} \lvert \times \lvert \mathcal{X} \lvert\) dimensional matrix called a transition matrix or stochastic matrix.

Properties

Finite Dimensional

\[p(X_k) = T p(X_{k-1}) = T^{k-1} X_1\]