Matrix transformations, inverses and eigenvalues
Introduzido| English | Significado |
|---|---|
| eigenvector/ˈaɪɡənvektə/ | A nonzero vector that a linear transformation maps to a scalar multiple of itself. |
| eigenvalues/ˈaɪɡənvæljuːz/ | The scale factors λ for which Ax = λx has a non-zero solution x. |
| determinant/dɪˈtɜːmɪnənt/ | Determinante |
| diagonalises | Turns a matrix into a diagonal matrix of eigenvalues, as in PᵀAP = D. |
| cofactors | The signed minors of the entries of a matrix, used to build its inverse. |
| characteristic equation/ˌkærɪktəˈrɪstɪk ɪˈkweɪʒn/ | The equation det(A − λI) = 0, whose roots are the eigenvalues of A. |
| orthogonal/ɔːˈθɒɡənl/ | Describes a square matrix whose columns are perpendicular unit vectors, so its inverse is its transpose. |
Ranking the web with an eigenvector, 1998
- In 1998, Sergey Brin and Larry Page at Stanford described PageRank, the method behind the first Google search. The rank of every web page is one entry of an eigenvector 特征向量 of a giant matrix of links.
- FP3 finds eigenvalues 特征值 and eigenvectors of 2 × 2 and 3 × 3 matrices, works with 3 × 3 determinants and inverses, and diagonalises 对角化 symmetric matrices.
3 × 3 determinants and inverses
- Expand along the top row: $\det\begin{pmatrix}1 & 2 & 0\\0 & 1 & 3\\2 & 0 & 1\end{pmatrix} = 1(1 - 0) - 2(0 - 6) + 0 = 13$. The volume scale factor is 13.
- The inverse is $\mathbf{A}^{-1} = \dfrac{1}{\det\mathbf{A}}\operatorname{adj}\mathbf{A}$, where the adjugate is the transposed matrix of cofactors 代数余子式. If $\det\mathbf{A} = 0$, there is no inverse.
Buscar $\det\begin{pmatrix}2 & 0 & 1\\1 & 3 & 0\\0 & 1 & 4\end{pmatrix}$.
Dificuldade 2/5 · Moedas 140 + pequeno bônus na primeira conclusão hoje
$2(12 - 0) - 0 + 1(1 - 0) = 25$.
A 3 × 3 matrix with two equal rows has an inverse.
Dificuldade 1/5 · Moedas 100 + pequeno bônus na primeira conclusão hoje
Its determinant is 0, so it is singular.
Eigenvalues and eigenvectors
- An eigenvector $\mathbf{x} \ne \mathbf{0}$ keeps its direction: $\mathbf{A}\mathbf{x} = \lambda\mathbf{x}$. The eigenvalue $\lambda$ is its stretch factor.
- Eigenvalues solve the characteristic equation 特征方程 $\det(\mathbf{A} - \lambda\mathbf{I}) = 0$.
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.
Find the eigenvalues of $\begin{pmatrix}5 & 2\\2 & 2\end{pmatrix}$.
Dificuldade 1/5 · Moedas 100 + pequeno bônus na primeira conclusão hoje
$\lambda^2 - 7\lambda + 6 = 0$. The diagonal entries are not the eigenvalues.
Worked example: a 2 × 2 matrix
Find the eigenvalues and eigenvectors of $\mathbf{A} = \begin{pmatrix}4 & 1\\2 & 3\end{pmatrix}$.
- Characteristic equation: $(4 - \lambda)(3 - \lambda) - 2 = \lambda^2 - 7\lambda + 10 = 0$, so $\lambda = 2$ or $5$.
- $\lambda = 5$: $\begin{pmatrix}-1 & 1\\2 & -2\end{pmatrix}\begin{pmatrix}x\\y\end{pmatrix} = \mathbf{0}$ gives $y = x$, eigenvector $\begin{pmatrix}1\\1\end{pmatrix}$.
- $\lambda = 2$: $\begin{pmatrix}2 & 1\\2 & 1\end{pmatrix}\begin{pmatrix}x\\y\end{pmatrix} = \mathbf{0}$ gives $y = -2x$, eigenvector $\begin{pmatrix}1\\-2\end{pmatrix}$. Check: $\mathbf{A}\begin{pmatrix}1\\-2\end{pmatrix} = \begin{pmatrix}2\\-4\end{pmatrix}$ ✓.
Which is an eigenvector of $\begin{pmatrix}5 & 2\\2 & 2\end{pmatrix}$ for $\lambda = 6$?
Dificuldade 1/5 · Moedas 100 + pequeno bônus na primeira conclusão hoje
$(5 - 6)x + 2y = 0$ gives $x = 2y$. The zero vector never counts.
Checking a 3 × 3 eigenvector
- You can test a given vector directly: for $\mathbf{B} = \begin{pmatrix}2 & 0 & 0\\0 & 3 & 4\\0 & 4 & -3\end{pmatrix}$, $\mathbf{B}\begin{pmatrix}0\\2\\1\end{pmatrix} = \begin{pmatrix}0\\10\\5\end{pmatrix} = 5\begin{pmatrix}0\\2\\1\end{pmatrix}$, so $\lambda = 5$.
- The eigenvalues of $\mathbf{B}$ are $2$, $5$ and $-5$; their sum, 2, equals the sum of the leading diagonal.
A favor $\mathbf{B} = \begin{pmatrix}2 & 0 & 0\\0 & 3 & 4\\0 & 4 & -3\end{pmatrix}$, $(0, 1, -2)$ is an eigenvector. Find its eigenvalue.
Dificuldade 2/5 · Moedas 140 + pequeno bônus na primeira conclusão hoje
$\mathbf{B}(0, 1, -2) = (0, 3 - 8, 4 + 6) = (0, -5, 10) = -5(0, 1, -2)$.
Diagonalising a symmetric matrix
- A symmetric matrix has perpendicular eigenvectors. Normalise them (length 1) and use them as the columns of $\mathbf{P}$. Then $\mathbf{P}$ is orthogonal 正交的 ($\mathbf{P}^{-1} = \mathbf{P}^{\mathrm{T}}$), and $\mathbf{P}^{\mathrm{T}}\mathbf{A}\mathbf{P} = \mathbf{D}$, a diagonal matrix of eigenvalues.
- $\mathbf{A} = \begin{pmatrix}2 & 1\\1 & 2\end{pmatrix}$ has $\lambda = 1, 3$ with $\tfrac{1}{\sqrt{2}}(1, -1)$ and $\tfrac{1}{\sqrt{2}}(1, 1)$, so $\mathbf{D} = \begin{pmatrix}1 & 0\\0 & 3\end{pmatrix}$.
For an orthogonal matrix $\mathbf{P}$, $\mathbf{P}^{-1} = \mathbf{P}^{\mathrm{T}}$.
Dificuldade 1/5 · Moedas 100 + pequeno bônus na primeira conclusão hoje
Its columns are perpendicular unit vectors, so $\mathbf{P}^{\mathrm{T}}\mathbf{P} = \mathbf{I}$.
Giving the zero vector, or mixing the column order. $\mathbf{x} = \mathbf{0}$ always satisfies $\mathbf{A}\mathbf{x} = \lambda\mathbf{x}$, so it is never an eigenvector. And in $\mathbf{P}^{\mathrm{T}}\mathbf{A}\mathbf{P} = \mathbf{D}$, the eigenvalues on the diagonal of $\mathbf{D}$ must be in the same order as their eigenvectors in $\mathbf{P}$.
Expand 3 × 3 determinants along a row; invert with the adjugate over the determinant. Eigenvalues solve $\det(\mathbf{A} - \lambda\mathbf{I}) = 0$, and each gives a non-zero eigenvector from $(\mathbf{A} - \lambda\mathbf{I})\mathbf{x} = \mathbf{0}$. A symmetric matrix is diagonalised by an orthogonal $\mathbf{P}$ of normalised eigenvectors: $\mathbf{P}^{\mathrm{T}}\mathbf{A}\mathbf{P} = \mathbf{D}$.
Find an orthogonal matrix $\mathbf{P}$ and a diagonal matrix $\mathbf{D}$ with $\mathbf{P}^{\mathrm{T}}\mathbf{A}\mathbf{P} = \mathbf{D}$ for $\mathbf{A} = \begin{pmatrix}5 & 2\\2 & 2\end{pmatrix}$.
Eigen-pairs, normalise, match order.