Matrix transformations, inverses and eigenvalues
Introduit| English | Sens |
|---|---|
| eigenvector/ˈaɪɡənvektə/ | A nonzero vector that a linear transformation maps to a scalar multiple of itself. |
| matrix/ˈmeɪtrɪks/ | A rectangular array of numbers arranged in rows and columns. |
| determinant/dɪˈtɜːmɪnənt/ | déterminant |
Ranking the web with an eigenvector, 1998
- In 1998 Sergey Brin and Larry Page described PageRank, the method behind early Google. They arranged the links between web pages in a huge matrix. The importance score of every page is an eigenvector 特征向量 of that matrix — a direction the matrix does not change.
- A matrix 矩阵 stores numbers in rows and columns. As a transformation it moves points; its determinant 行列式 gives the area scale factor, and its eigenvectors show the directions it only stretches.
PageRank scores are an eigenvector of the web's link matrix.
Difficulté 1/5 · 100 pièces + un petit bonus lors de la première réussite aujourd'hui
A direction the matrix does not change.
Matrix arithmetic
- Multiply row by column: $\begin{pmatrix}a & b\\ c & d\end{pmatrix}\begin{pmatrix}x\\y\end{pmatrix} = \begin{pmatrix}ax + by\\ cx + dy\end{pmatrix}$. In general $AB \ne BA$.
- Inverse of a $2\times2$ matrix: $A^{-1} = \dfrac{1}{ad - bc}\begin{pmatrix}d & -b\\ -c & a\end{pmatrix}$, if $\det A = ad - bc \ne 0$.
(1 2; 3 4) times the column (1, 1) equals
Difficulté 1/5 · 100 pièces + un petit bonus lors de la première réussite aujourd'hui
Rows times column.
Find det (5 2; 3 4).
Difficulté 2/5 · 140 pièces + un petit bonus lors de la première réussite aujourd'hui
20 − 6.
Transformations
- The columns of $A$ are the images of $\begin{pmatrix}1\\0\end{pmatrix}$ and $\begin{pmatrix}0\\1\end{pmatrix}$. Areas scale by $|\det A|$.
Matrix transformations, inverses and eigenvalues
A 2×2 matrix [[a,b],[c,d]] has determinant ad-bc and inverse [[d,-b],[-c,a]]/(ad-bc) if the determinant is nonzero
Compare the model with the worked case and explain one change.
A shape of area 4 is transformed by (2 1; 0 3). Find its new area.
Difficulté 2/5 · 140 pièces + un petit bonus lors de la première réussite aujourd'hui
4 × 6.
Eigenvalues and eigenvectors
- $A\mathbf v = \lambda\mathbf v$: the matrix only stretches $\mathbf v$ by $\lambda$. Solve $\det(A - \lambda I) = 0$ for $\lambda$.
- Applying $A$ many times, the largest eigenvalue's direction dominates — the idea behind PageRank and long-run Markov chains.
What are the eigenvalues of (2 1; 0 3)?
Difficulté 1/5 · 100 pièces + un petit bonus lors de la première réussite aujourd'hui
(2 − λ)(3 − λ) = 0.
Worked example: the parallelogram
$A = \begin{pmatrix}2 & 1\\ 0 & 3\end{pmatrix}$ sends $(1, 0) \to (2, 0)$ and $(0, 1) \to (1, 3)$.
- Determinant: $2 \times 3 - 1 \times 0 = 6$: the unit square becomes a parallelogram of area 6.
- Inverse: $A^{-1} = \tfrac16\begin{pmatrix}3 & -1\\ 0 & 2\end{pmatrix}$.
- Eigenvalues: $\det\begin{pmatrix}2 - \lambda & 1\\ 0 & 3 - \lambda\end{pmatrix} = (2 - \lambda)(3 - \lambda) = 0$, so $\lambda = 2, 3$. For $\lambda = 3$: $-x + y = 0$, eigenvector $(1, 1)$. Check: $A(1, 1) = (3, 3)$. ✓
Which is an eigenvector for λ = 3?
Difficulté 1/5 · 100 pièces + un petit bonus lors de la première réussite aujourd'hui
A(1, 1) = (3, 3).
Markov chains
- A transition matrix $T$ has columns summing to 1; the state after $n$ steps is $T^n\mathbf s_0$. The steady state is the eigenvector with $\lambda = 1$.
Matrix multiplication is not commutative. $AB$ means "apply $B$ first, then $A$" for transformations, and usually differs from $BA$. Keep the order.
For matrices, AB always equals BA.
Difficulté 1/5 · 100 pièces + un petit bonus lors de la première réussite aujourd'hui
Order matters.
Multiply matrices row by column; order matters. A $2\times2$ matrix's columns are the images of the basis vectors, its determinant is the area factor, and its inverse exists when $\det \ne 0$. Eigenvalues solve $\det(A - \lambda I) = 0$; the dominant eigenvector gives long-run behaviour.
Each week 80% of gym members stay and 20% leave; 10% of non-members join. Write the transition matrix and find the long-run proportion of members.
Matrix, steady state, meaning.