Linear Transformations

Consider a map (function) \( T \) between \( \mathcal{U} \) & \( \mathcal{V} \):

Two blobs labeled U and V with arrow T from U to V mapping point x to y equals T of x
Figure 1
A mapping diagram. On the left, a blue closed curve (blob) is labeled \( \mathcal{U} \) and contains a red dot labeled \( x \). On the right, a slightly smaller blue blob is labeled \( \mathcal{V} \) and contains a red dot labeled \( y = T(x) \). Between the two blobs, a straight black arrow points from left to right and is labeled \( T \) above it. A curved red arrow starts at the point \( x \) inside the left blob, travels rightward across the gap, and ends with an arrowhead at the point \( y = T(x) \) inside the right blob, illustrating that the map \( T \) sends the element \( x \) of the domain to its image in the target.

\( \mathcal{U} \) : domain; \( \mathcal{V} \): target

\( y = T(x) \) is called the image of \( x \)

\( x \in T^{-1}\{y\} \) is called a pre-image of \( y \)

Linear Transformations

Consider a map (function) \( T \) between \( \mathcal{U} \) & \( \mathcal{V} \):

Blob U with point x, arrow T to blob V with point y equals T of x
Figure 2
A hand-drawn mapping diagram. On the left is a blue closed blob labeled \( \mathcal{U} \) containing a red dot labeled \( x \). On the right is a blue closed blob labeled \( \mathcal{V} \) containing a red dot labeled \( y = T(x) \). A straight black arrow labeled \( T \) points from the left region to the right region across the top middle. A curved red arrow runs from the point \( x \) in \( \mathcal{U} \) to the point \( y = T(x) \) in \( \mathcal{V} \), showing that \( T \) sends \( x \) to its image.

\( T \) is onto (surjective) if any \( y \in \mathcal{V} \) is the image of some \( x \in \mathcal{U} \),

i.e. for any \( y \in \mathcal{V} \), there is an \( x \in \mathcal{U} \) such that \( y = T(x) \)

Linear Transformations

Consider a map (function) \( T \) between \( \mathcal{U} \) & \( \mathcal{V} \):

Blob U containing point x, arrow T to blob V containing point y equals T of x
Figure 3
A mapping diagram. On the left, a blue closed curve labeled \( \mathcal{U} \) contains a red dot labeled \( x \). On the right, a blue closed curve labeled \( \mathcal{V} \) contains a red dot labeled \( y = T(x) \). A straight black arrow labeled \( T \) points from the left region toward the right region, and a curved red arrow runs from the point \( x \) in \( \mathcal{U} \) to the point \( y = T(x) \) in \( \mathcal{V} \).

\( T \) is one-to-one (injective)

if \( T(x_1) = T(x_2) \implies x_1 = x_2 \)

ie. \( x_1 \ne x_2 \implies T(x_1) \ne T(x_2) \)

i.e. if \( T^{-1}\{y\} \) is non-empty, \( T^{-1}\{y\} \) has only one element. unique element.

Linear Transformations

Consider a map (function) \( T \) between \( \mathcal{U} \) & \( \mathcal{V} \):

Blobs U and V joined by arrow T to the right and arrow T inverse to the left, with x and y equals T of x
Figure 4
A mapping diagram. On the left is a blue closed curve labeled \( \mathcal{U} \) containing a red dot labeled \( x \), with a second red label beneath it reading \( x = T^{-1}(y) \). On the right is a blue closed curve labeled \( \mathcal{V} \) containing a red dot labeled \( y = T(x) \). A black arrow labeled \( T \) points from the left region to the right region, and below it a black arrow labeled \( T^{-1} \) points from the right region back to the left region. Two red curved arrows connect the points directly: one from \( x \) to \( y \) and one from \( y \) back to \( x \).

If \( T \) is both one-to-one and onto, then \( T \) has an inverse \( T^{-1} : \mathcal{V} \longrightarrow \mathcal{U} \).

  1. \( T^{-1} \circ T(x) = x \), ie. \( T^{-1} \circ T = \) identity on \( \mathcal{U} \)
  2. \( T \circ T^{-1}(y) = y \), ie. \( T \circ T^{-1} = \) identify on \( \mathcal{V} \)

Linear Transformation between \( \mathbb{R}^n \) & \( \mathbb{R}^m \)

Consider \( \mathbb{R}^n \xrightarrow{\;T\;} \mathbb{R}^m \), lin. transf.

Given by \( T(X) = AX \)

\( X \in \mathbb{R}^n \)   \( AX \in \mathbb{R}^m \), \( A - m \times n \) matrix

\( T \) is onto if for any \( Y \in \mathbb{R}^m \), there is an \( X \in \mathbb{R}^n \) such that \( AX = Y \),

ie. \( AX = Y \) is solvable for any \( Y \)

i.e \( \mathrm{Rank}(A) = m \) ( no zero rows in REF of \( A \))

Linear Transformation between \( \mathbb{R}^n \) & \( \mathbb{R}^m \)

Consider \( \mathbb{R}^n \xrightarrow{\;T\;} \mathbb{R}^m \), lin. transf.

Given by \( T(X) = AX \)

\( X \in \mathbb{R}^n \)   \( AX \in \mathbb{R}^m \), \( A - m \times n \) matrix

\( T \) is one-to-one if \( AX_1 = AX_2 \implies X_1 = X_2 \)

ie. if \( AX = Y \) is solvable, there is only one soln.

ie. if \( AX = 0 \implies X = 0 \), ie. \( \mathrm{Null}(A) = \{\vec{0}\} \)

ie. \( \mathrm{Nullity}(A) = 0 \), (no free variables)

Linear Transformation between \( \mathbb{R}^n \) & \( \mathbb{R}^m \)

Consider \( \mathbb{R}^n \xrightarrow{\;T\;} \mathbb{R}^m \), lin. transf.

Given by \( T(X) = AX \)

\( X \in \mathbb{R}^n \)   \( AX \in \mathbb{R}^m \), \( A - m \times n \) matrix

\( T \) is onto and one-to-one, if

\( \mathrm{Rank}(A) = m = n \), \( (\implies A \) is necessarily square\( ) \)

For any \( Y \), \( AX = Y \) is uniquely solvable.

Inverse of a Square Matrix A

\( \mathbb{R}^n \;\underset{T^{-1}}{\overset{T}{\rightleftarrows}}\; \mathbb{R}^n \)

\( T(X) = AX, \qquad T^{-1}(Y) = A^{-1} Y \)

\( T^{-1} \circ T(X) = X \) for any \( X \) \( \implies \) \( A^{-1} A = I \)

\( T \circ T^{-1}(Y) = Y \) for any \( Y \) \( \implies \) \( A A^{-1} = I \)

Def: \( B \) is called the inverse of \( A \), denoted by \( A^{-1} \), if \( AB = I \) & \( BA = I \)

Inverse of a Square Matrix A

  1. For a square matrix \( A, {}^{(n \times n)} \)

    \( A \) is onto \( \iff A \) is one-to-one

    \( \Updownarrow \)

    \( \mathrm{Rank}(A) = n \) (no zero rows in REF\((A)\)) \( \iff \) \( \mathrm{Nullity}(A) = 0 \) (no free variables)

  2. If \( BA = I \), then \( AB = I \) ( and hence \( B = A^{-1} \))

  3. \( A^{-1} \) is unique:

    if \( AB = BA = I \), \( \forall \) \( AC = CA = I \), then \( B = C \).

Inverse of a Square Matrix A

(4) If \( A^{-1} \) & \( B^{-1} \) exist, then \( (AB)^{-1} = B^{-1} A^{-1} \)

Diagram of three copies of R to the n linked by B and A forward, inverses back, with AB and AB inverse arcs
Figure 5
A commutative-style diagram showing three copies of \( \mathbb{R}^n \) arranged left to right. Between the first and second copy there is a pair of black arrows: a rightward arrow labeled \( B \) above, and a leftward arrow labeled \( B^{-1} \) below. Between the second and third copy there is a similar pair: a rightward arrow labeled \( A \) above and a leftward arrow labeled \( A^{-1} \) below. A long red curved arrow arcs over the top from the first \( \mathbb{R}^n \) to the third \( \mathbb{R}^n \), labeled \( AB \). A second long red curved arrow arcs underneath from the third \( \mathbb{R}^n \) back to the first \( \mathbb{R}^n \), labeled \( (AB)^{-1} \).

(5) \( (A^{T})^{-1} = (A^{-1})^{T} \)

\[ \left( A^{T}(A^{-1})^{T} = (A^{-1}A)^{T} = I^{T} = I \right) \]

Inverse of a Square Matrix \( A^{n \times n} \)

How to find \( A^{-1} \) (if it exists)

\( \Updownarrow \) \( \mathrm{Rank}(A) = n \)

\[ \left[\begin{array}{c|c} A & I \end{array}\right] \xrightarrow{\;\text{elementary row op.}\;} \cdots \longrightarrow \left[\begin{array}{c|c} I & B \end{array}\right] \]

identity (\( I \) in \( [A \mid I] \)),   identity (\( I \) in \( [I \mid B] \)),   \( A^{-1} \) (\( B \))