☺ Matrix Algebra – By Jorge ☺ Page 1

Consider the system of equations: Spanning Set

[ ] [ ] [ ]
−3x1 x 2−2x 3=−5 | | |
Coefficient Matrix Augmented Matrix Span A1 , A2 ,⋯ , An =c1 A1c 2 A2 ⋯c n An

[ ]
| | |

[ ]
1 -2 1 1 -2 1 2
2 1 -1 2 1 -1 1 A span is all the linear combinations of vectors A1 to An
1 1 -2 1 1 -2 -5 n
Tip: A span can be equal ℝ as long as it can represent
Matrix Multiplication every single member in it, it is a subspace otherwise.
m×n× n× z=m×z

Compatible?
Note: A span is also a subspace
New Dimensions Subspace
n
Non-Singular Matrix Definition: describes a subset of ℝ
Must meet 3 criteria while maintaining condition in order
A is singular if the only solution to Ax is 0
to be a subspace

[ ][]
x1 0 (1) Must contain the 0 vector

[ ]
| | | | x2 0 (2) Must be closed under addition. Is x  y in
A1 A2 A3 ⋯ An x3 = 0 subspace?
| | | | ⋮ ⋮ (3) Must be closed under scalar multiplication. Is
xn 0 c x in subspace (c is a constant)
where Ai are the columns of A. Note: Subspace is usually in this notation, but can have
Augmented Matrix: [ A1 A2 A3⋯ An∣0 ] other forms, like a spanning set

{ [] }
Key Points x1
(1) Singular means that there is a solution other than 0 W = x : x= x 2 ; <condition>
vector. ⋮
(2) Also, non-singular matrices are square, meaning xn
that they n x n
(3) When row reduced, the matrix simplifies to the Proving Linear Independence
If vectors { x
1 ,x 2 ,⋯, 
x n } are linearly independent, then
identity matrix
Transpose of a Matrix in the equation c 1  x1 c 2 
x 2⋯c n  x n=0 , constants
Tip: Flip the columns and rows c 1=c 2=c 3=⋯=c n =0 , they have to be equal to 0.
Dimension

[ ]
− A1 −
T dim W =number of linearly independent vectors

[ ]
| | | | − A1 − nullity=dim Nullspace W 
A1 A2 A3 ⋯ An = − A2 − Rank W = Number of pivot columns and rows
| | | | ⋮ # of Columns of W =Rank W nullity W 
− An − Magnitude or distance of a vector is ∥q∥= q T q
Properties Note: All definition above are numbers only, not vectors!
T T T Products
1)  AB  = A  B T
T T T Dot Product of 2 vectors u⋅v =u v

2)  AC  =C A

[] []
T T
u1 v1
3)  A  = A Cross Product of 2 vectors. u= u2 v = v2
T T T
4)  A A = A A u v
3 3

[ ]
u 2 v 3−u3 v 2
Note: A matrix is known to be symmetric if it is equal to its
T Then u×v= u3 v 1−u1 v 3
transpose. In other words, it's symmetric when A = A .
u1 v 2−u2 v 1
Hint – only applies to n×n matrices
Note: Cross product lets you get an orthogonal vector to
that of u and v.
 ☺ Matrix Algebra – By Jorge ☺ Page 2

Orthogonality Basis
Definition: Orthogonal vectors are the set of vectors whose Definition: A non-zero spanning set that is linearly
dot product is equal to 0. In other words u⋅v =0 independent.
n
Note: Getting the dot product of an orthogonal vector to Note: In order to be a basis in ℝ , the set can have no
2
itself such as  u =∥u∥ =uT⋅u≠0 .
u⋅ more than n vectors and no less than 1.

Gram-Schmidt Process Note:To prove a basis, you must prove that the vectors are
Allows us to calculate an orthogonal basis linearly independent.
u 1=w 1 Homogeneous System
T
u1 w2 Definition: An m×n augmented matrix that is equal to 0.
u 2=w2 − T u1 Ex. [ A| b ] where b= 0 , thus always has 0 as one
u1 u1
solution.
u T1 w 3 uT2 w 3
u 3=w 3− T u 1− T u 2 Possible Scenarios:
u1 u1 u2 u2 (1) If m < n – Infinitely Many Solutions
T
i−1
u w (2) If m = n – Unique or Infinitely Many Solutions
u i=wi−∑ kT i u k where 2≤i≤ p Vector Form of a Solution
k=1 u k u k
Let A be an augment matrix
Normalization
Definition: A normal vector has a magnitude or distance of
just 1 unit.
To calculate, divide the vector by its magnitude, i.e.
q
[ 1 0 -1 0 -1 -2 0
A= 0 1 2 0 1 2 0 the corresponding
0 0 0 1 1 1 0
solutions are:
]
∥q∥ x 1= x3 x 52 x 6
Normal Equation x 2=−2 x 3−x 5−2 x6
Definition: the equation that allows you to find the most x 4 =−x5 −x 6
T * T
approximate solution (w*) to b: A A w = A b thus the corresponding Vector Form is:

[] [] [] []
x1 1 1 2
Another process that helps you find w* x2 -2 -1 -2
vT u x = x 3 : x3 1 x 0 x 0
p
w *=∑ T i u i
5 6
x4 0 -1 -1
i=1 u i u i x4 0 1 0
where p is number of vectors in A, v is the provided x6 0 0 1
inconsistent solution and u i are the corresponding
orthogonal basis members . Nullspace
Definition: It is the solution set (or subspace ) that makes a
Note: w* is known to be the least-square solution matrix be equal to the 0 vector. In other words:
N  A={x : Ax=0 , x ∈ ℝ n}
Note: Because w* will be the most approximate solution to
b, then ∥Aw *−v∥≤∥Ax−v∥ where x is any other vector Hint: The basis for a matrix that is equal to the 0 vector
is the same basis for the nullspace
Linear Transformations
Definition: A linear transformation pertains to a function Inverse of a Matrix
where T  x  = A x . In other words, matrix A can Definition: The matrix is invertible if it is n×n and such
represent all algebraic combinations, once multiplied with that: A−1 A= A A−1= I
x, that could ultimately represent any linear transformation To calculate, augment the matrix with the identity and row
T. reduce to the identity:
In order to be a linear transformation the following 2
criteria must be met:

[ ] [ ]
1 2 3 1 0 0 1 0 0 54 -23 -7
(1) T  x  T  y  =T  x  y  Addition 2 5 4 0 1 0 0 1 0 -16 7 2

(2) c T  x  =T  c x  Scalar Multiplication 1 -1 10 0 0 1 0 0 1 -7 3 1