โ›“๏ธ Markov Chain

Reinforcement Learning / Environment

The Markov property states that the future only depends on the present. Markov chains use this property to model dynamic systems that transition between states; the next state depends only on the current state.

A chain consists of a set of states, probabilities of initializing at each state, and a matrix of transition probabilities across states. The following is a graphical example.

Formally, the chain consists of the state space and transition operator . For states , we wish to capture the probability

using the transition operator. If we let and , then

describes how the probability distribution of states changes over time via our transitions.

Content by William Liang, written in Obsidian.
Thank you to all the educators who made these notes possible.