LINEAR ALGEBRA Week 2 1.7 Linear Independence 1.8 Introduction to Linear Transformations 1.9 The Matrix of a Linear Transformation Neil V. Budko
December 15, 2016
Neil V. Budko
Lay 1.7, 1.8, 1.9
1.7 Linear Independence
Neil V. Budko
Lay 1.7, 1.8, 1.9
1.7 Linear Independence
Linearly dependent and linearly independent vectors Linear dependence and solutions of homogeneous systems When the set of vectors is linearly dependent/independent?
Neil V. Budko
Lay 1.7, 1.8, 1.9
Definition of linear independence
DEFINITION: A set of vectors {v1 , v2 , . . . , vp } in Rn is linearly independent if the vector equation: x1 v1 + x2 v2 + · · · + xp vp = 0 has only the trivial solution (all weights are zero x1 = x2 = · · · = xp = 0). The set {v1 , v2 , . . . , vp } in Rn is linearly dependent if there exists weights {c1 , c2 , . . . , cp , not all equal to zero, such that: c1 v1 + c2 v2 + · · · + cp vp = 0
Neil V. Budko
Lay 1.7, 1.8, 1.9
Connection to homogeneous equations
A set of vectors {v1 , v2 , . . . , vp } in Rn is linearly independent if the homogeneous matrix equation: Ax = 0, where A = [v1 v2 . . . vp ] has only the trivial solution x = 0. The set {v1 , v2 , . . . , vp } in Rn is linearly dependent if the above homogeneous matrix equation has a nontrivial solution x 6= 0.
Neil V. Budko
Lay 1.7, 1.8, 1.9
Example 1.7.1 Let 1 4 2 v1 = 2 , v2 = 5 , v3 = 1 3 6 0 (a) Are these vectors linear independent? (b) If they are not, show one linear dependence relation between them. SOLUTION: (a) Write down the augmented matrix of the equivalent homogeneous matrix equation A = [v1 v2 v3 0] and row-reduce it: 1 4 2 0 1 4 2 0 1 4 2 0 2 5 1 0 ∼ 0 −3 −3 0 ∼ 0 −3 −3 0 3 6 0 0 0 −6 −6 0 0 0 0 0
Neil V. Budko
Lay 1.7, 1.8, 1.9
Example 1.7.1 (continued) Clearly x3 is a free variable. Hence, the homogeneous system Ax = 0 has a nontrivial solution x 6= 0 for each value of x3 . (b) To show a linear row-reduction of the 1 4 2 0 −3 −3 0 0 0
dependence relation perform the complete augmented matrix: 0 1 4 2 0 1 0 −2 0 0 ∼ 0 1 1 0 ∼ 0 1 1 0 0 0 0 0 0 0 0 0 0
Now, we see that x1 = 2x3 , x2 = −x3 , and x3 is free. Choose any value for x3 , say x3 = 5. Then, x1 = 10 and x2 = −5. These are the possible values of the weights in the linear dependence relation between v1 , v2 , and v3 . Hence, we can write this relation down as: 10v1 − 5v2 + 5v3 = 0 Neil V. Budko
Lay 1.7, 1.8, 1.9
Linear independence of matrix columns
The columns of matrix A are linearly independent if and only if the homogeneous equation Ax = 0 has only the trivial solution x = 0. (If there exists a nontrivial solution x 6= 0, then the columns of A are linearly dependent.)
Neil V. Budko
Lay 1.7, 1.8, 1.9
Example 1.7.2 Determine if the columns of the matrix A are linearly independent. 0 1 4 A = 1 2 −1 5 8 0
SOLUTION: 0 1 4 0 1 2 −1 0 1 2 −1 0 1 2 −1 0 ∼ 0 1 4 0 ∼ 0 1 4 0 5 8 0 0 5 8 0 0 0 0 13 0 There are no free variables, and the only possible solution is x = 0. Hence, the columns are linearly independent.
Neil V. Budko
Lay 1.7, 1.8, 1.9
Seeing linear dependence by inspection
Set of one vector {v} is linear dependent if and only if v = 0, since in that case cv = 0 for any c 6= 0. Set of two vectors {v1 , v2 } is linear dependent if and only if one of these vectors is a scalar multiple of another. For example, let v2 = av1 , then the linear dependence relation is: c1 v1 + c2 v2 = c1 v1 + c2 av1 = (c1 + c2 a)v1 = 0 Hence, we can always find c1 and c2 both not equal to zero, such that c1 + c2 a = 0, i.e. pick any c2 6= 0 and set c1 = −c2 a.
Neil V. Budko
Lay 1.7, 1.8, 1.9
Example 1.7.3
Consider the following sets for linear dependence: 3 6 3 6 (a) v1 = , v2 = ; (b) v1 = , v2 = . 1 2 2 2
SOLUTION: These are sets of two vectors, so we may use the inspection technique. (a) Clearly v2 = 2v1 , and the set is linearly dependent. (b) There does not exist a constant a, such that v2 = av1 . Hence, the set is linearly independent.
Neil V. Budko
Lay 1.7, 1.8, 1.9
Sets of two or more vectors
Theorem. A set S = {v1 , . . . , vp } of two or more vectors is linearly dependent if and only if at least one of these vectors is a linear combination of the other vectors. In fact, if S is linearly dependent and v1 6= 0, then there is a vector vj in this set which is a linear combination of the preceding vectors v1 , v2 , . . . , vj−1 . Remark. There may be vectors in S that are not linear combinations of the preceding vectors. There may also be vectors in S that are not linear combinations of other vectors (from S) at all.
Neil V. Budko
Lay 1.7, 1.8, 1.9
Linear dependence and the size of the set
Theorem. If the set S = {v1 , . . . , vp } in Rn contains more vectors than the length of its vectors, i.e., if p > n, then the set S is linearly dependent.
Neil V. Budko
Lay 1.7, 1.8, 1.9
Linear dependence and the zero vector
Theorem. Every set containing the zero vector 0 is linearly dependent.
Neil V. Budko
Lay 1.7, 1.8, 1.9
Example 1.7.6 Consider the following sets for linear dependence: −2 3 1 2 3 4 2 0 1 4 −6 (a) 7 , 0 , 1 , 1 (b) 3 , 0 , 1 (c) 6 , −9 6 9 5 8 0 0 8 10 15
SOLUTION: (a) This is a set of 4 vectors each length 3. It is linearly dependent since 4 > 3. (b) This set is linearly dependent because it contains the zero vector. (c) This is a linearly independent set of two vectors, since they are not scalar multiples of each other.
Neil V. Budko
Lay 1.7, 1.8, 1.9
1.8 Introduction to Linear Transformations
Neil V. Budko
Lay 1.7, 1.8, 1.9
1.8 Introduction to Linear Transformations
Transformations, their domain, codomain, and range Analysis of matrix transformations Linear transformations
Neil V. Budko
Lay 1.7, 1.8, 1.9
Domain, codomain, and range
1
A transformation T : Rn → Rm from Rn to Rm is a rule T that assigns to each vector x ∈ Rn a vector T (x) ∈ Rm
2
The set Rn is called the domain of the transformation T
3
The set Rm is called the codomain of the transformation T
4
For a given vector x ∈ Rn the vector T (x) ∈ Rm is called the image of x under the action of T . The set of all images T (x) is called the range of T
Neil V. Budko
Lay 1.7, 1.8, 1.9
Domain, codomain, and range
Fact: The range of T can be smaller than the codomain of T .
Neil V. Budko
Lay 1.7, 1.8, 1.9
Matrix multiplication as a transformation
Ax = b, and Au = 0 1 1 5 4 −3 1 3 4 −3 1 3 1 4 = 0 = and 0 8 2 0 5 1 −1 2 0 5 1 1 3 1
Neil V. Budko
Lay 1.7, 1.8, 1.9
Matrix transformations
Let x ∈ Rn and let T (x) be computed as Ax, where A ∈ Rm×n . Then, The domain is Rn The codomain is Rm The range is a subset of Rm consisting of all possible linear combinations of the columns of A, i.e., the Span of the columns of A
Neil V. Budko
Lay 1.7, 1.8, 1.9
Example 1.8.1 (analysis of matrix transformations)
Let
1 −3 3 3 2 5 ,u = A= 3 , b = 2 , c = 2 , −1 −1 7 −5 5 and define the transformation T : R2 → R3 as T (x) = Ax. (a) Find T (u), the image of u under the transformation T . (b) Find a vector x whose image under T is b. (c) Is there more than one vector whose image under T is b? (d) Determine if c is in the range of T .
Neil V. Budko
Lay 1.7, 1.8, 1.9
Example 1.8.1 (continued) (a) To find the image of u simply compute Au: 1 −3 5 2 5 Au = 3 = 1 −1 −1 7 −9 (b) To find pre-image of b, solve Ax = b for x: 3 1 −3 x1 3 5 = 2 x2 −5 −1 7 Row-reduction of 1 −3 3 5 −1 7
the augmented matrix: 3 1 −3 3 1 −3 3 2 ∼ 0 14 −7 ∼ 0 14 −7 ∼ −5 0 4 −2 0 0 0 Neil V. Budko
Lay 1.7, 1.8, 1.9
Example 1.8.1 (continued)
Row-reduction continued: 1 −3 3 1 0 1.5 ∼ 0 1 −0.5 ∼ 0 1 −0.5 0 0 0 0 0 0 The solution is the vector x=
1.5 −0.5
(c) This pre-image vector is unique (since there are no free variables). Hence, there are no other x whose image is b.
Neil V. Budko
Lay 1.7, 1.8, 1.9
Example 1.8.1 (continued)
(d) To find if c is in the range of T , perform row-reduction of the matrix A augmented by the vector c: 1 −3 3 1 −3 3 1 −3 3 3 5 2 ∼ 0 14 −7 ∼ 0 14 −7 −1 7 5 0 4 8 0 0 10 There is a pivot in the right-hand-side column (last row). Hence, the system is inconsistent, i.e., there is no x such that Ax = c. Therefore, c is not in the range of A, i.e., not in the range of T .
Neil V. Budko
Lay 1.7, 1.8, 1.9
Example 1.8.2 (projection)
The following matrix-vector product defines the projection transformation of points on R3 to the (x1 , x2 )-plane in R3 : x1 1 0 0 x1 0 1 0 x2 = x2 0 x3 0 0 0
Neil V. Budko
Lay 1.7, 1.8, 1.9
Example 1.8.3 (shear) This matrix-vector product is a shear transformation in (x1 , x2 )-plane: 1 3 x1 x1 + 3x2 = 0 1 x2 x2
Check the transformation of each corner.
Neil V. Budko
Lay 1.7, 1.8, 1.9
General linear transformations
Definition. A transformation T is linear if: 1
T (u + v) = T (u) + T (v) for all u and v in the domain of T .
2
T (cu) = cT (u) for all scalars c and all u in the domain of T .
Fact. Every matrix transformation is a linear transformation: A(u + v) = Au + Av,
Neil V. Budko
A(cu) = cAu
Lay 1.7, 1.8, 1.9
Beware of transformations that look linear
Consider the affine transformation: T (x) = Ax + b Is it linear? It sure looks linear. Wait, checking... T (u + v) = A(u + v) + b = Au + Av + b while T (u) + T (v) = Au + b + Av + b No, it is not!
Neil V. Budko
Lay 1.7, 1.8, 1.9
Superposition principle
If T is a linear transformation, then: 1
T (0) = 0.
2
T (cu + dv) = cT (u) + dT (v) for all scalars c and d, and all u and v in the domain of T .
The general superposition principle: T (c1 v1 + c2 v2 + · · · + cp vp ) = c1 T (v1 ) + c2 T (v2 ) + · · · + cp T (vp ) Given a system that performs a linear transformation (linear system), a linear combination of inputs will result in a linear combination of outputs with the same weights (same linear combination).
Neil V. Budko
Lay 1.7, 1.8, 1.9
Contraction and dilation
Multiplication by a scalar r is a linear transformation T (x) = r x (prove it). It is called contraction, if 0 ≤ r < 1, and dilation, if r > 1.
Neil V. Budko
Lay 1.7, 1.8, 1.9
Example 1.8.5 (rotation) This is an example of a counterclockwise rotation transformation by π/2 radian in (x1 , x2 )-plane: 0 −1 x1 −x2 = 1 0 x2 x1
Neil V. Budko
Lay 1.7, 1.8, 1.9
1.9 The Matrix of a Linear Transformation
Neil V. Budko
Lay 1.7, 1.8, 1.9
1.9 The Matrix of a Linear Transformation
Standard matrix for the linear transformation Geometric linear transformations on R2 Existence and uniqueness questions
Neil V. Budko
Lay 1.7, 1.8, 1.9
The identity matrix The identity matrix In is a n-by-n matrix with ones on the main diagonal. For example a 3-by-3 identity matrix I3 is 1 0 0 I3 = 0 1 0 0 0 1 The main property of I is In x = x, for all x ∈ Rn The columns of the identity matrix are denoted by ei , e.g. in R3 , 1 0 0 e1 = 0 , e2 = 1 , e1 = 0 0 0 1 Hence, every x ∈ Rn can be represented as: x = In x = x1 e1 + x2 e2 + · · · + xn en . Neil V. Budko
Lay 1.7, 1.8, 1.9
The standard matrix of a linear T (x)
Theorem. Let T : Rn → Rm be a linear transformation. Then, there exists a unique matrix A such that T (x) = Ax, for all x ∈ Rn . In fact, A is an m-by-n matrix whose j-th column is the vector T (ej ) ∈ Rm obtained by acting with T on the vector ej (j-th column of the n-by-n identity matrix): A = [T (e1 ) T (e2 ) . . . T (en )]
Neil V. Budko
Lay 1.7, 1.8, 1.9
The standard matrix - Proof The claim is that every linear transformation T (x) between Rn and Rm is a matrix transformation Ax with a special kind of m-by-n matrix A. Indeed: T (x) = T (In x) = T (x1 e1 + x2 e2 + · · · + xn en ) = T (x1 e1 ) + T (x2 e2 ) + · · · + T (xn en ) = x1 T (e1 ) + x2 T (e2 ) + · · · + xn T (en ) x1 x2 = [T (e1 )T (e2 ) . . . T (en )] . = Ax .. xn Fact: Every (discrete) linear transformation is a matrix transformation and vise versa. The columns of the standard matrix can be obtained by acting with T on the columns of the unit matrix. Neil V. Budko
Lay 1.7, 1.8, 1.9
Example 1.9.2 (finding the standard matrix)
Find the standard matrix of the dilation transformation T (x) = 3x for x ∈ R2 . SOLUTION: We compute the columns of A by acting with T on the columns of the unit matrix I2 : 1 3 0 0 T (e1 ) = 3e1 = 3 = , T (e2 ) = 3e2 = 3 = 0 0 1 3 3 0 A= 0 3
Neil V. Budko
Lay 1.7, 1.8, 1.9
Example 1.9.3 (standard matrix of the rotation) Find the standard matrix of the transformation T (x) that rotates x ∈ R2 by ϕ radians in the positive direction. SOLUTION: It can be proven geometrically that this is a linear transformation. Hence, we only need to consider the action of T on the columns of the unit matrix in R2 . These columns are the basis vectors e1 = i and e2 = j of the 2D Cartesian system. Hence, from elementary geometry: x-component of the rotated e1 vector cos(ϕ) T (e1 ) = = y -component of the rotated e1 vector sin(ϕ) x-component of the rotated e2 vector − sin(ϕ) T (e2 ) = = y -component of the rotated e2 vector cos(ϕ)
Neil V. Budko
Lay 1.7, 1.8, 1.9
Example 1.9.3 (continued)
The standard matrix of this rotation transformation is: cos(ϕ) − sin(ϕ) A= sin(ϕ) cos(ϕ)
Neil V. Budko
Lay 1.7, 1.8, 1.9
Geometrical transformations on R2
Neil V. Budko
Lay 1.7, 1.8, 1.9
Geometrical transformations on R2
Neil V. Budko
Lay 1.7, 1.8, 1.9
Geometrical transformations on R2
Neil V. Budko
Lay 1.7, 1.8, 1.9
Geometrical transformations on R2
Neil V. Budko
Lay 1.7, 1.8, 1.9
Geometrical transformations on R2
Neil V. Budko
Lay 1.7, 1.8, 1.9
Geometrical transformations on R2
Neil V. Budko
Lay 1.7, 1.8, 1.9
Onto transformations
Definition. A mapping T : Rn → Rm is said to be onto Rm if each b in Rm is the image of at least one x in Rn . How to check: The transformation is onto, if its standard matrix has pivots in each row.
Neil V. Budko
Lay 1.7, 1.8, 1.9
One-to-one transformations Definition. A mapping T : Rn → Rm is said to be one-to-one if each b in Rm is the image of at most one x in Rn .
“At most one” is tricky! If some b is not in the range of T , then it is not an image of any x, but it is still “an image of at most one x”. Think about it. How to check: The transformation is one-to-one, if its standard matrix has pivots in each column.
Neil V. Budko
Lay 1.7, 1.8, 1.9
Example 1.9.4 (onto and one-to-one stuff) Given the standard matrix 1 −4 8 1 A = 0 2 −1 3 0 0 0 5 is the transformation defined by this matrix onto, one-to-one? SOLUTION: The transformation is between R4 and R3 . A is already in the echelon form. Pivots are present in all rows. Hence, Ax = b is consistent, i.e., there exists a solution x ∈ R4 for each b ∈ R3 . (each b is an image of some x). Thus, T is onto. However, not all columns have pivots (one free variable). Hence, the solution of Ax = b is not unique (more than one x has b as an image). Thus, T is not one-to-one.
Neil V. Budko
Lay 1.7, 1.8, 1.9
One-to-one and the homogeneous equation
Theorem. A linear transformation T : Rn → Rm is one-to-one if and only if the equation T (x) = 0 has only the trivial solution.
Neil V. Budko
Lay 1.7, 1.8, 1.9
Onto and one-to-one for the standard matrix Theorem. Let A be the standard matrix of the linear transformation T : Rn → Rm . Then: T is onto Rm if and only if the columns of A span Rm T is one-to-one if and only if the columns of A are linearly independent To analyze a linear transformation: 1
Check if T : Rn → Rm is indeed a linear transformation
2
Get the standard matrix A of T by working with T on the columns of In
3
To check onto property: get the echelon form of A, check if each row has a pivot
4
To check one-to-one property: inspect columns of A or get the echelon form of A and check if each column has a pivot
Neil V. Budko
Lay 1.7, 1.8, 1.9
Example 1.9.5 (weird notation) Let T be defined as: T (x) = T (3x1 + x2 , 5x1 + 7x2 , x1 + 3x2 ) where the input vector is in R2 with components x1 and x2 , and the output vector (see comas inside T ) is in R3 . Is T onto, one-to-one? SOLUTION: Deducing the standard matrix by inspection: ? ? 3x1 + x2 3 1 x x T (x) = 5x1 + 7x2 = ? ? 1 = 5 7 1 x2 x2 x1 + 3x2 ? ? 1 3
Neil V. Budko
Lay 1.7, 1.8, 1.9
Example 1.9.5 (continued) So the standard matrix is 3 1 A = 5 7 1 3 Applying the Theorem. (a) T : R2 → R3 is onto R3 if and only if the columns of A span R3 . This is not possible, since we only have two columns, and there will be at most two pivots, while we need three. So T is not onto R3 (b) T : R2 → R3 is one-to-one if and only if the columns of A are linearly independent. Since we have two columns, we only need to check if they are scalar multiples of each other. They are not. Hence, they are linearly independent, and T is one-to-one.
Neil V. Budko
Lay 1.7, 1.8, 1.9