Why introduce a matrix instead of just reading the transition diagram?

Conditions

• The Markov chain has fixed one-step transition probabilities.
• State vectors are represented in a consistent order.

Reasoning, step by step

1. Recognize that reading the diagram is manageable for one step but becomes inefficient for many steps.
2. Encode the one-step transition probabilities into a transition matrix.
3. Use matrix-vector multiplication to compute the next state vector.
4. Iterate the matrix multiplication to efficiently find states far into the future.

Example

The speaker asks how to go from $S_0$ to $S_1$, then from $S_1$ to $S_2$, or $S_2$ to $S_100$, and says there is a process to do those manipulations efficiently using a matrix.

Common misconceptions

• Thinking the diagram alone is enough for many steps.
• Believing that the matrix formalism is already necessary just to compute $S_1$, whereas the video explicitly says that for the first step one could read off the answer from the diagram.

Watch the explanation

Explore next

• In this video, what do the rows and columns of the transition matrix represent?
• How do you compute the next state vector from the current one?