Matrices Modeling Contexts
| English | Chinese | Pinyin |
|---|---|---|
| state vector | 状态向量 | zhuàng tài xiàng liàng |
| state | 状态 | zhuàng tài |
| transition matrix | 转移矩阵 | zhuǎn yí jǔ zhèn |
Predicting the future with a matrix
- Every year, some people move from the city to the suburbs, and some move back.
- Given this year's populations, can you predict next year's?
- A matrix can store the movement rates and do the prediction for you.
- This is how matrices model populations, markets, weather, and more.
State vectors: the current picture
- Collect the current amounts into a state vector 状态向量 — one entry per category.
- For a city and its suburbs: $\begin{bmatrix}\text{city}\\ \text{suburb}\end{bmatrix}$, say $\langle 600, 400\rangle$ (thousands).
- Each state 状态 is one thing being tracked; the vector is a snapshot in time.
- Next year's snapshot will be a new state vector.
A state vector in this kind of model holds…
A state vector lists how much is in each category right now — one entry per state.
The transition matrix: the rules
- A transition matrix 转移矩阵 stores the fraction moving between every pair of states.
- Each number is a rate: e.g. $5\%$ of city dwellers move to the suburbs each year.
- The diagram below shows two states with the movement rates on the arrows.
- Those same rates become the entries of the matrix.
The matrix that stores the rates of movement between states is called the ____ matrix.
The transition matrix holds the fraction moving from each state to each other state.
Multiplying the transition matrix by the current state vector predicts the next state.
One matrix-times-vector step advances the model by one time period.
Stepping forward in time
- Multiply the transition matrix by the current state vector to get next year's.
- One matrix-times-vector = one year forward.
- Want two years? Multiply again. Ten years? Multiply ten times.
- The model marches forward, one step per multiplication.
A good use of a matrix?
Matrices model transformations, systems and transitions. Sort each task.
To predict the state two periods ahead, you…
Each multiplication advances one period, so two periods need two multiplications (or multiply by the matrix squared).
Select all true statements about matrix models.
The same transition matrix is reused each period — you do not redraw it. The other three are correct.
The long run
- Repeated steps often settle toward a steady state that stops changing.
- At the steady state, as many people leave the city as arrive — populations hold constant.
- Matrices let you find this long-run balance without simulating every single year.
- The same idea powers web-page ranking, genetics, and epidemic models.
Set the model up consistently: decide once whether the state vector is a column multiplied on the right of the matrix, and make each column's (or row's) rates add to $1$ so no one is lost or invented. A mismatched setup silently gives nonsense predictions.
City $= 600$, suburb $= 400$ (thousands). Each year $5\%$ of city → suburb, $2\%$ of suburb → city.
- Leaving the city: $0.05 \times 600 = 30$; arriving: $0.02 \times 400 = 8$.
- New city $= 600 - 30 + 8 = 578$; new suburb $= 400 + 30 - 8 = 422$.
- The matrix does exactly this arithmetic in a single multiplication.
A state vector holds the current amount in each state, and a transition matrix stores the rates of movement between them. Multiplying the matrix by the state vector steps the model one period forward; repeating it predicts the long-run behaviour. This is how matrices model real-world change.