MIT 18.06SC Linear Algebra Session 1.1 - 1.5

10 Mar 2018- Zhiqiang Zang


Hmmm… I’m learning MIT 18.06SC Linear Algebra at MIT OpenCourseWare.

OK, here we go!

Unit I: \(A\boldsymbol{x} = \boldsymbol{b}\) and the Four Subspaces

Session 1.1: The Geometry of Linear Equations

We have a system of equations:

\[\left\{ \begin{aligned} 2x - y &= 0 \\ -x + 2y &= 3 \end{aligned} \right .\]

Row Picture

Line \(2x - y = 0\) and line \(-x + 2y = 0\) intersects at the point \((1, 2)\), so \((1, 2)\) is the solution of the system of equations.

Maybe I should draw a X-Y coordinates here >_>

Column Picture

We rewrite the system of linear equations as a single equation:

\[x\begin{bmatrix}2 \\ -1\end{bmatrix} + y\begin{bmatrix}-1 \\ 2\end{bmatrix} = \begin{bmatrix}0 \\ 3\end{bmatrix}\]

We see \(x\) and \(y\) as scalars of column vectors: \(\boldsymbol{v_1} = \begin{bmatrix}2 \\ -1\end{bmatrix}\) and \(\boldsymbol{v_2} = \begin{bmatrix}-1 \\ 2\end{bmatrix}\), and the sum \(x\boldsymbol{v_1} + y\boldsymbol{v_2}\) is called a linear combination of \(\boldsymbol{v_1}\) and \(\boldsymbol{v_2}\).

Geometrically, we can find one copy of \(\boldsymbol{v_1}\) added to two copies of \(\boldsymbol{v_2}\) just equals the vector \(\begin{bmatrix}0 \\ 3\end{bmatrix}\). Then the solution should be \(x = 1, y =2\).

I will add a figure when time is available >_>

Matrix Picture

We rewrite the equations in our example as a compact form,

\[A\boldsymbol{x} = \boldsymbol{b},\]

that is

\[\begin{bmatrix}2 & -1 \\ -1 & 2\end{bmatrix}\begin{bmatrix}x \\ y\end{bmatrix} = \begin{bmatrix}0 \\ 3\end{bmatrix}\]

Matrix Multiplication

\[\begin{aligned} \begin{bmatrix}2 & -1 \\ -1 & 2\end{bmatrix} \begin{bmatrix}1 \\ 2\end{bmatrix} = 1\begin{bmatrix}2 \\ -1\end{bmatrix} + 2\begin{bmatrix}-1 \\ 2\end{bmatrix} = \begin{bmatrix}0 \\ 3\end{bmatrix} \end{aligned}\]

A matrix times by a vector is just a linear combination of the column vectors of the matrix.


Session 1.2: An Overview of Key Ideas

Vectors

Let us take linear combinations of vectors.

Matrices

The product of a matrix and a vector is a combination of the columns of the matrix.

Subspaces

All combinations of column vectors creates a subspace. The subspaces of \(\mathbb{R}^3\) are:

Conclusion


Session 1.3: Elimination with Matrices

Method of Elimination

We have an example \(A\boldsymbol{x} = \boldsymbol{b}\),

\(A = \begin{bmatrix}1 & 2 & 1 \\ 3 & 8 & 1 \\ 0 & 4 & 1\end{bmatrix}\) and \(\boldsymbol{b} = \begin{bmatrix}2 \\ 12 \\2\end{bmatrix}\).

Steps of Elimination:

\[A = \begin{bmatrix}1 & 2 & 1 \\ 3 & 8 & 1 \\ 0 & 4 & 1\end{bmatrix} \rightarrow \begin{bmatrix}1 & 2 & 1 \\ 0 & 2 & -2 \\ 0 & 4 & 1\end{bmatrix} \rightarrow U = \begin{bmatrix}1 & 2 & 1 \\ 0 & 1 & -2 \\ 0 & 0 & 5\end{bmatrix}\] \[\boldsymbol{b} = \begin{bmatrix}2 \\ 12 \\ 2\end{bmatrix} \rightarrow \cdots \rightarrow \begin{bmatrix}2 \\ 6 \\ -10\end{bmatrix}\]

Thus, we can easily solve the systems of equations, \(\begin{bmatrix}x \\ y \\ z\end{bmatrix} = \begin{bmatrix}2 \\ 1 \\ -2\end{bmatrix}\).

Elimination Matrices

The product of a matrix (\(3 \times 3\)) and a column vector (\(3 \times 1\)) is a column vector (\(3 \times 1\)) that is a linear combination of the columns of the matrix.

The product of a row vector (\(1 \times 3\)) and a matrix (\(3 \times 3\)) is a row vector (\(1 \times 3\)) that is a linear combination of the rows of the matrix.

For example,

\[\begin{bmatrix}1 & 0 & 0 \\ -3 & 1 & 0 \\ 0 & 0 & 1\end{bmatrix} \begin{bmatrix}1 & 2 & 1 \\ 3 & 8 & 1 \\ 0 & 4 & 1\end{bmatrix} = \begin{bmatrix}1 & 2 & 1 \\ 0 & 2 & -2 \\ 0 & 4 & 1\end{bmatrix}.\]

Multiplying on the left by a permutation matrix exchanges the rows of a matrix, while multiplying on the right exchanges the columns. For example,

\[P = \begin{bmatrix}0 & 1 & 0 \\ 1 & 0 & 0 \\ 0 & 0 & 1\end{bmatrix}.\]

\(P\) is a permutation matrix and the first and second rows of the matrix \(PA\) are the second and first rows of the matrix \(A\).

Note, matrix multiplication is associative but not commutative.

Inverses

We have a matrix:

\[E_{21} = \begin{bmatrix}1 & 0 & 0 \\ -3 & 1 & 0 \\ 0 & 0 & 1\end{bmatrix}\]

which subtracts \(3\) times row \(1\) from row \(2\). To “undo” this operation we must add \(3\) times row \(1\) to row \(2\) using the inverse matrix:

\[E_{21}^{-1} = \begin{bmatrix}1 & 0 & 0 \\ 3 & 1 & 0 \\ 0 & 0 & 1\end{bmatrix}.\]

In fact, \(E_{21}^{-1}E_{21} = I\).


Session 1.4: Multiplication and Inverse Matrices

Four and a half ways we see matrix multiplication

We have \(AB = C\). \(A\) is an \(m \times n\) matrix and \(B\) is an \(n \times p\) matrix, then \(C\) is an \(m \times p\) matrix. We use \(c_{ij}\) to denote the entry in row \(i\) and column \(j\) of matrix \(C\) and the same denotation applies to \(a_{ij}\) and \(b_{ij}\).

Row times column

\[c_{ij} = \sum_{k=1}^n a_{ik}b_{kj}\]

Columns

The product of matrix \(A\) and column \(j\) of matrix \(B\) equals column \(j\) of matrix \(C\). This tells us that the columns of \(C\) are combinations of columns of \(A\).

\[A \begin{bmatrix} | & | & | \\ column 1 & column 2 & column 3 \\ | & | & | \end{bmatrix} =\begin{bmatrix} | & | & | \\ A(column 1) & A(column 2) & A(column 3) \\ | & | & | \end{bmatrix}\]

Rows

The product of row \(i\) of matrix \(A\) and matrix \(B\) equals row \(i\) of matrix \(C\). So the rows of \(C\) are combinations of rows of \(B\).

\[\left[ \begin{matrix} --- & row 1 & --- \\ --- & row 2 & --- \\ --- & row 3 & --- \end{matrix} \right] B =\left[ \begin{matrix} --- & (row 1)B & --- \\ --- & (row 2)B & --- \\ --- & (row 3)B & --- \end{matrix} \right]\]

Column times row

\[AB = \sum_{k=1}{n} \begin{bmatrix}a_{1k} \\ \vdots \\ a_{mk}\end{bmatrix} \begin{bmatrix}b_{k1} & \cdots & b_{kp}\end{bmatrix}\]

note: Here I fixed a typo in the lecture summary (PDf) of this session: \(b_{kp}\) instead of the original \(b_{kn}\).

Blocks

\[\begin{bmatrix}A_1 & A_2 \\ A_3 & A_4 \end{bmatrix} \begin{bmatrix}B_1 & B_2 \\ B_3 & B_4 \end{bmatrix} =\begin{bmatrix} A_1B_1+A_2B_3 & A_1B_2+A_2B_4 \\ A_3B_1+A_4B_3 & A_3B_2+A_4B_4 \end{bmatrix}\]

Inverses

If \(A\) is singular or not invertible,

then A does not have an inverse,

and we can find some non-zero vector \(\boldsymbol{x}\) for which \(A\boldsymbol{x} = \boldsymbol{0}\)

Gauss-Jordan Elimination

\(E \left[ \begin{array}{c|c} A & I \end{array} \right] =\left[ \begin{array}{c|c} I & E \end{array} \right]\) If \(EA = I\), then \(E = A^{-1}\).


Session 1.5: Factorization into \(A = LU\)

Inverse of a product

\[(AB)^{-1} = B^{-1}A^{-1}\]

Transpose of a product

\[(AB)^{T} = B^{T}A^{T}, \quad (A^{T})^{-1} = (A^{-1})^{T}\]

\(A = LU\)

We can use elimination to convert \(A\) into an upper triangular matrix \(U\), that is \(EA = U\), and further we can also convert this to a factorization \(A = LU\) in which \(L = E^{-1}\).

For example, in a three dimensional case, if \(E_{32}E_{31}E_{21}A = U\) then \(A=E_{21}^{-1}E_{31}^{-1}E_{32}^{-1}U = LU\). Suppose \(E_{31}\) is the identity matrix and \(E_{32}\) and \(E_{21}\) are as shown below:

\[\begin{array}{cccc} E_{32} & E_{21} & & E \\ \begin{bmatrix}1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & -5 & 1\end{bmatrix} & \begin{bmatrix}1 & 0 & 0 \\ -2 & 1 & 0 \\ 0 & 0 & 1\end{bmatrix} & = & \begin{bmatrix}1 & 0 & 0 \\ -2 & 1 & 0 \\ 10 & -5 & 1\end{bmatrix} \end{array}.\]

Here \(L = E^{-1} = E_{21}^{-1}E_{32}^{-1}\):

\[\begin{array}{cccc} E_{21}^{-1} & E_{32}^{-1} & & L \\ \begin{bmatrix}1 & 0 & 0 \\ \underline{2} & 1 & 0 \\ 0 & 0 & 1\end{bmatrix} & \begin{bmatrix}1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & \underline{5} & 1\end{bmatrix} & = & \begin{bmatrix}1 & 0 & 0 \\ \underline{2} & 1 & 0 \\ 0 & \underline{5} & 1\end{bmatrix} \end{array}\]

Notice the \(0\) in row three column one of \(L\), where \(E\) had a \(10\). The factorization \(A = LU\) is preferable to the statement \(EA = U\) because the combination of row subtractions does not have the effect on \(L\) that it did on \(E\).

If there are no row exchanges, the multipliers from the elimination matrices are copied directly into \(L\).

Cost of elimination

If we define a typical operation is to multiply one row and then subtract it from another, then the total number of operations needed to factor \(n \times n\) \(A\) into \(LU\) is on the order of \(n^3\):

\[1^2 + 2^2 + \cdots + (n - 1)^2 + n^2 = \sum_{i = 1}^{n}i^2 \approx \int_0^n x\,\mathrm{d}x = \frac{1}{3}n^3.\]

While we’re factoring \(A\) we’re also operating on \(\boldsymbol{b}\). That costs about \(n^2\) operations, which is hardly worth counting compared to \(1/3n^3\).

Row exchanges

The inverse of any permutation matrix \(P\) is \(P^{-1} = P^{T}\).

There are \(n!\) different ways to permute the rows of an \(n \times n\) matrix (including the permutation that leaves all rows unfixed) so there are \(n!\) permutation matrices. These matrices form a multiplicative group.


This work is built upon the materials on MIT OCW.


LICENSE:

Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.