Matrix Notations

 Concepts
 A scalar is a numerical value
 Example: 4, 100, 4.5, 5, 2/3, …
 Variables for scalars use symbols x, y, z, …
 A vector is a sequence of values
 Row vector: (2, 6, 3), (50, 4.5, ½, 54), …

6
3
0
 Column vector: 4.2 , −5 , …
6 1/2
 Variables for vectors use symbols x, y, z, …
 Without mentioned otherwise, vector variables always denote
column vectors
 In text composition, we write “…x = (2,1,3)T…”
2
 Matrix Notations
 Concepts
 A matrix is block of values with rows and columns
 Examples:
2 0.4
 A 3x2 matrix: 8 3
−3 0

3 7 0 1 3
2 1 4 5 9
 A 4x5 matrix:
5 −2 7 7 2
1 0 2 3 1
 Variables for matrixes use symbols X, Y, Z, …
 For any matrix A, the entry at the ith row and jth column is
denoted as Aij (or A(i, j)): A24 = 5, A32 = -2
3
 Matrix Notations
 Concepts
 A row vector with n values can be viewed as a 1xn matrix
 examples
 Row vector (2, 6, 3) viewed as 1x3 matrix

 Row vector (50, 4.5, ½, 54) viewed as 1x4 matrix

 A Column vector with n values can be viewed as nx1 matrix

 Examples
3
 Column vector 4.2 viewed as 3x1 matrix
6
6
0
 Column vector −5 viewed as 4x1 matrix
1/2
 A 1 x 1 matrix can be written simply as a scale: (a) = a 4
 Matrix Notations
 Matrixes can be represented in different ways
 In the most general case, an m x n matrix is represented as
𝑎11 ⋯ 𝑎1𝑛
A= ⋯ ⋯ ⋯
𝑎𝑚1 ⋯ 𝑎𝑚𝑛

5
 Matrix Addition/Subtraction
 Addition and subtraction between two matrices
 Operands must have same shape
 Operation performed on each matching element pair
 Example:
2 5 0 10 2 15
 8 3 + 1 5 = 9 8
1 0 2 0 3 0

4 1 2 1 5 2 3 −4 0
 - =
3 6 3 2 6 7 1 0 −4

6
 Multiplication with Scaler
 Multiplication between matrix and scalar
 Multiply the scalar to all the elements in the matrix
 Example
2 5 4 10
 2 8 3 = 16 6
1 0 2 0

1 5 2 5 25 10
 5=
2 6 7 10 30 35

7
 Matrix Multiplication
 Matrix multiplication AB
 Column count of A must match the row count of B
 Let A be a pxq matrix, and B be a qxr matrix, then AB = C
where
 C is a pxr matrix
 Cij = sum of the products of the matching elements in the ith row
of A and jth column of B
𝑎11 ⋯ 𝑎1𝑞 𝑏11 ⋯ 𝑎1𝑟
 Let 𝐀 = ⋯ ⋯ ⋯ , 𝐁= ⋯ ⋯ ⋯
𝑎𝑝1 ⋯ 𝑎𝑝𝑞 𝑎𝑞1 ⋯ 𝑎𝑞𝑟

σ𝑞𝑖=1 𝑎1𝑖 𝑏𝑖1 ⋯ σ𝑞𝑖=1 𝑎1𝑖 𝑏𝑖𝑟

 Then 𝐀𝐁 = ⋯ ⋯ ⋯
𝑞 𝑞
σ𝑖=1 𝑎𝑝𝑖 𝑏𝑖1 ⋯ σ𝑖=1 𝑎𝑝𝑖 𝑏𝑖𝑟
8
 Matrix Multiplication
 Example:

1 10
1 3 2 8 25
 1 5 =
0 2 1 2 0 4 10

2x3 3x2 2x2

1 10
 0, 2, 5 1 5 = (12, 10)
2 0
1x3 1x2
3x2

9
 Matrix Multiplication
2
 1 3 2 0 =8 (called inner product)
3
2 2 6 4
 0 1 3 2 = 0 0 0 (called outer product)
3 3 9 6

1 3 2 1 2

0 2 1 4 5
2x3 2x2

Not the same, operation undefined!

10
 Transpose
 Transpose
 The transpose of pxq matrix A is AT = B where
 B is a qxp matrix
 The ith row of A is identical to the ith column of B (or equivalently, the ith
column of A is identical to the ith row of B)
 A more concise definition would be:
 AT = B if Aij = Bji

𝑎11 ⋯ 𝑎1𝑞 𝑎11 ⋯ 𝑎𝑝1

 Let 𝐀 = ⋯ ⋯ ⋯ , then 𝐀T = ⋯ ⋯ ⋯
𝑎𝑝1 ⋯ 𝑎𝑝𝑞 𝑎1𝑞 ⋯ 𝑎𝑝𝑞
 Or let 𝐀 = 𝐀 𝟏 𝐀 𝟐 ⋯ 𝐀 𝐪 where Ai is px1 column vector

𝐀𝐓𝟏
 Then 𝐀𝐓 =
𝐀𝐓𝟐 where 𝐀𝐓𝐢 is 1xp row vector
⋮
𝐀𝐓𝐪
11
 Transpose
 Examples
1
𝑇
 1 2 3 = 2
1x3
3
3x1
𝑇
2
 0 = 2 0 3
3
1x3
3x1

𝑇 1 0
1 3 2
 = 3 2
0 2 1
2 1
2x3 3x2

12
 Transpose
𝑇
1 2 3 1 4 7 0
4 5 6 = 2 5

7 8 9 8 0
3 6 9 0
0 0 0

4x3 3x4

 Note: for any 1 x 1 matrix (a), we have

 𝑎 𝑇= 𝑎
 This is a special case of symmetric matrix (detail omitted)

13
 Determinant of Square Matrix
 For any square matrix A, its determinant is a scalar,
denoted as |A|
 The determinants of the matrices with the first few lowest
shapes are calculated as follows:
 |(a)| = a
𝑎 𝑏
 = 𝑎𝑑 − 𝑏𝑐
𝑐 𝑑
𝑎11 𝑎12 𝑎13
 𝑎21 𝑎22 𝑎23 = 𝑎11 𝑎22 𝑎33 + 𝑎21 𝑎32 𝑎13 + 𝑎31 𝑎23 𝑎12
𝑎31 𝑎32 𝑎33
− 𝑎13 𝑎22 𝑎31 − 𝑎23 𝑎32 𝑎11 − 𝑎33 𝑎21 𝑎12

 Note: matrices not in square shapes have no determinants

14
 Determinant of Square Matrix
 Examples
 |3| = 3
3 2
 = 34 − 21 = 10
1 4
1 2 3
 2 1 0
2 4 6

= 116 +243 + 202 − 312 − 041 − 622

= 30 – 30 = 0

15
 Determinant of Square Matrix
 A matrix with non-zero determinant is said to be in full rank,
otherwise not
3 2
 Matrices (2) and are in full rank
1 4

1 2 3
 2 1 0 is not in full rank
2 4 6
1 2 3
Proof: 2 1 0 = 1*1*6 + 2*4*3 + 2*0*2 – 3*1*2 – 0*4*1 – 6*2*2
2 4 6
= 6 + 24 + 0 – 6 – 0 -24 = 0

16
 Identity Matrix
 An identify matrix is a square matrix where the elements in its
main diagonal are all 1s, and the elements not in its main
diagonal are all 0s
 Examples of identity matrices

1 0 0
1 0
 1 , , 0 1 0 ,…
0 1
0 0 1

 Note: the determinant of any identity matrix is 1

 An nxn identity matrix is denoted as Inxn

17
 Identity Matrix
 For any nxn matrix A, AInxn = InxnA = A
 This can be easily verified
𝑎 𝑏 1 0 𝑎 𝑏
 =
𝑐 𝑑 0 1 𝑐 𝑑
1 0 𝑎 𝑏 𝑎 𝑏
 =
0 1 𝑐 𝑑 𝑐 𝑑
𝑎11 𝑎12 𝑎13 1 0 0 𝑎11 𝑎12 𝑎13
 𝑎21 𝑎22 𝑎23 0 1 0 = 𝑎21 𝑎22 𝑎23
𝑎31 𝑎32 𝑎33 0 0 1 𝑎31 𝑎32 𝑎33
1 0 0 𝑎11 𝑎12 𝑎13 𝑎11 𝑎12 𝑎13
 0 1 0 𝑎21 𝑎22 𝑎23 = 𝑎21 𝑎22 𝑎23
0 0 1 𝑎31 𝑎32 𝑎33 𝑎31 𝑎32 𝑎33
 .....
18
 Inverse Matrix
 If a square matrix A is in full rank, then it has an inverse
matrix, denoted as A-1, defined as follows:
 AA-1 = A-1A = I
 An analogy of basic algebra: for any value a  0, aa-1 = a-1a = 1
 The inverse matrix for 1x1 and 2x2 matrices are computed as
follows:
 (a)-1 = (1/a)

−1 𝑑 −𝑏
𝑎 𝑏 −𝑐 𝑎 1 𝑑 −𝑏
 = 𝑎 𝑏 =
𝑐 𝑑 𝑎𝑑−𝑏𝑐 −𝑐 𝑎
𝑐 𝑑
 Verify the above matrices fit the definition for the inverse matrix
 Calculations for the inverse matrices of larger sizes are omitted

19
 About Numpy Module for Python
 Numpy is a module that must be installed separately from
Python interpreter
 The installation procedure varies depending on the
platform (refer to the idle help manual for detail)
 NumPy’s main object is the homogeneous
multidimensional array
 It is a table of elements (usually numbers), all of the same
type, indexed by a tuple of positive integers
 Dimensions are called axes
 In this course, we never use more than two dimensions
 A two dimensional array is a matrix
20
 About Numpy Module
 NumPy’s array class is called ndarray
 Note this is different from the standard Python array class
 It provides more powerful functionalities

21
 Attributes of ndarray class
• ndarray.ndim
• the number of axes (dimensions) of the array.
• ndarray.shape
• the dimensions of the array, which is a tuple of integers indicating
the size of the array in each dimension.
• For a matrix with n rows and m columns, shape is (n,m).
• ndarray.size
• the total number of elements of the array.
• This is equal to the product of the elements of shape.
• ndarray.dtype
• an object describing the type of the elements in the array

22
 How to create an ndarray
 Use array() function: supply a >>> a.ndim
sequence as argument 1
>>> a.shape
>>> import numpy as np
(3,)
>>> a = np.array([2,3,4])
>>> a
Note: you must supply a sequence:
array([2, 3, 4])
>>> a.dtype
dtype('int64’) >>> a = np.array(1,2,3,4) # WRONG
>>> b = np.array([1.2, 3.5, 5.1]) >>> a = np.array([1,2,3,4]) # RIGHT
>>> b.dtype
dtype('float64')
23
 How to create an ndarray (cont.)
>>> a = np.array([[1.5,2,3], [4,5,6]]) >>> b = np.array([[2,3,4]])
>>> a >>> b
array([[ 1.5, 2. , 3. ], array([[2, 3, 4 ]])
[ 4. , 5. , 6. ]]) >>> b.ndim
Note. This is the matrix: 2
1.5 2 3 >>> b.shape
4 5 6
(1,3)
>>> a.ndim
2
>>> a.shape
(2,3)
24
 How to create an ndarray (cont.)
>>> a=np.array([2,3,4]) >>> b=np.array([[2,3,4]]) >>> c=np.array([[2],[3],[4]])
>>> c
>>> a >>> b
array([[2],
array([2, 3, 4]) array([[2, 3, 4]]) [3],
>>> a.ndim >>> b.ndim [4]])
1 2 >>> c.ndim
2
>>> a.shape >>> b.shape
>>> c.shape
(3,) (1,3) (3,1)

a 3x1 matrix:
a three dimensional row a 1x3 matrix:
vector:
2
[2, 3, 4]
[2, 3, 4] 3
4
25
 Basic Operations
 Arithmetic operators on arrays apply elementwise

>>> a = np.array( [20,30,40,50] )

>>> a
array([20,30,40,50])
>>> b = np.array([0,1,2,3])
>>> b
array([0, 1, 2, 3])
>>> c = a-b
>>> c
array([20, 29, 38, 47])
26
 Basic Operations (cont.)
>>> a = np.array([[0, 1, 2], [3, 4, 5]]) >>> b = np.array( [[20,30],[40,50]] )
>>> a >>>b
array([[0, 1, 2], array([[20,30],
[3, 4, 5]]) [40,50]])
>>> a**2 >>> 10*np.sin(b)
array([[0, 1, 4], array([[9.12945251, -9.88031624],
[9, 16, 25]]) [7.4511316 , -2.62374854]])
>>> b<35
array([ [True, True],
[False, False] ])
27
 Aggregation Operations
 Defined as methods for ndarray class
>>> a = np.array([[1,2,3],[4,5,6]]) >>> a.sum(axis=0)
>>> a array([5,7,9])
array([[1,2,3], >>> a.sum(axis=1)
[4,5,6]]) array([6,15])
>>>a.sum() >>>a.max(axis=1))
21 array([3,6])
>>>a.max() >>>a.min(axis=0)
6 array([1,2,3])
>>>a.min()
1 The results are all 1D array

28
 Math Functions
 All applied on individual entries of the matrix
>>> B = np.array([0,1,2])
>>> B
array([0, 1, 2])
>>> np.exp(B) #exponent for base e = 2.7182818284….
array([ 1. , 2.71828183, 7.3890561 ])
>>> np.sqrt(B)
array([ 0. , 1. , 1.41421356])
>>> C = np.array([2., -1., 4.])
>>> np.add(B, C)
array([ 2., 0., 6.])

29
 Indexing and Slicing
 One-dimensional arrays can be indexed, sliced and
iterated over, much like lists and other Python sequences.
>>> a = np.arange(10)**3
>>> a
array([ 0, 1, 8, 27, 64, 125, 216, 343, 512, 729])
>>> a[2]
start stop (exclusive)
8
>>> a[2:8:2] #interpretation: a[2:8:2]
array([ 8, 64, 216]) step
>>> a[:6:2] = -1000 #same as a[0:6:2]
>>> a
array([-1000, 1, -1000, 27, -1000, 125, 216, 343, 512, 729])

30
 For multi-dimensional arrays
 Have one index per axis, given in a tuple separated by commas
>>> a = np.array([[0,1,2,3], >>> a[2,3]
[10,11,12,13], 23
[20,21,22,23], >>> a[0:4, 1]
[30,31,32,33], array([1, 11, 21, 31])
[40,41,42,43]]) >>> a[:, 1]
>>> a array([1, 11, 21, 31, 41])
array([[ 0, 1, 2, 3], >>> a[1:4, 0:3]
[10, 11, 12, 13], array([[10, 11, 12],
[20, 21, 22, 23], [20, 21, 22],
[30, 31, 32, 33], [30, 31, 32]])
[40, 41, 42, 43]])

31
 Iterating
 Over all the designated values
>>> a = np.array([1,8,27,64,125,216]) >>> for i in b:
>>> for i in a: print(i)
print(i**(1/3)) [1,2,3]
1 [4,5,6]
2 >>> for i in b:
3 for j in i[0:2]:
4 print(j**2)
5 1
6 4
>>> b = np.array([[1,2,3], 16
[4,5,6]]) 25
32
 Stacking together different arrays
>>> a = np.array([[1,2,3], [4,5,6]]) >>> c = np.vstack((a, b))
>>> a >>> c
array([[1,2,3], array([[1,2,3],
[4,5,6]]) [4,5,6],
>>> b = np.array([[6,5,4],[3,2,1]]) [6,5,4],
>>> b [3,2,1]])
array([[6,5,4], >>> d = np.hstack((a,b))
[3,2,1]]) array([[1,2,3, 6,5,4],
[4,5,6, 3,2,1]])

33
 Repeat
>>> np.repeat(3, 4) >>> np.repeat(a, 2, axis = 1)
array([3,3,3,3]) array([[1,1,2,2,3,3],
>>> np.repeat([1,2], 4) [4,4,5,5,6,6]])
array([1,1,1,1,2,2,2,2]) >>> np.repeat(a, [1,2], axis = 0)
>>> a = np.array([[1,2,3],[4,5,6]]) array([[1,2,3],
>>> a [4,5,6],
array([[1,2,3], [4,5,6]])
[4,5,6]]) >>> np.repeat(a, [1,2,2], axis=1)
>>> np.repeat(a, 2, axis = 0) array([[1,2,2,3,3],
array([[1,2,3], [4,5,5,6,6]])
[1,2,3], >>>np.repeat(a, [1,2,1],axis=0)
[4,5,6], Error…
[4,5,6]])

34
 Compare matrices with same shape
>>>a = np.array([[2,4],[1,2]] >>>d = a > b
>>>b = np.array([[0,4],[1,3]]) >>>d
>>>a array([[True,False],
array([[2,4], [False,False]])
[1,2]])
>>>b
array([[0,4],
[1,3]])
>>>c = a == b
>>>c
array([[False,True],
[True,False]])

35
 Compare matrices with difference shapes
>>>a = np.array([[2,4,2],[1,2,3]] >>>d = np.array([[4],[1]])
>>>a >>>d
array([[2,4,2], array([[4],
[1,2,3]]) [1]])
>>>b = np.array([2,2,3]) >>>a == d
>>>b array([[False,True,False],
array([2,2,3]) [True,False,False]])
>>>c = a == b >>>f = np.array([4,1])
>>>c >>>a == f
array([[True,False,False], False
[False,True,True]]) >>>a > f
error…

36
 Compare matrices with difference shapes
>>>a = np.array([[2,4,2],[1,2,3]] >>>a + 1 < 4
>>>a array([[True,False,True],
array([[2,4,2], [True,True,False]])
[1,2,3]])
>>>a == 2
array([[True,False,True],
[False,True,False]])
>>>a > 2
array([[False,True,False],
[False,False,True]])

37
 Count non-zeros in a matrix
>>>a = np. array([[1,0,2], >>>np.count_nonzero(a, axis=0)
[0,1,0]]) array([1,1,1])
>>>a >>>np.count_nonzero(a,axis=1)
array([[1,0,2], array([2,1])
[0,1,0]])
>>>np.count_nonzero(a)
3

38
 Retrieve locations for non-zeros in a matrix
>>>a = np. array([[1,0,2], >>>b = np.array([1,0,0,2,3])
[0,1,0]]) >>>b
>>>a array([1,0,0,2,3])
array([[1,0,2], >>>np.nonzero(b)
[0,1,0]]) (array([0,3,4]))
>>>np.nonzero(a)
(array([0,0,1]), array([0,2,1]))

39
 More for operations on non-zeros
>>>a = np. array([[True,False,True], >>>np.nonzero(a)
[False,True,False]]) (array([0,0,1]), array([0,2,1]))
>>>a >>>b = np.array([True,False,True])
array([[True,False,True], >>>b
array([True,False,True])
[False,True,False]])
>>>np.nonzero(b)
>>>np.count_nonzero(a)
(array([0,2]))
3
>>>np.count_nonzero(a, axis=0)  Combining operations on non-
array([1,1,1]) zeros and matrix comparisons,
>>>np.count_nonzero(a,axis=1) you can retrieve counts and
locations for any values and
array([2,1]) their duplicates in a matrix

40
 More for Operations on Non-Zeros
>>>a = np.array([1,2,5,2,2,5,0])
 How to find the locations of all the occurrences of 2?
>>>b = a == 2
>>>b
array([False,True,False,True,True,False,False])
>>>c = np.nonzero(b)
>>>c
array([1,3,4])
 [1,3,4] gives the locations of occurrences of 2 in a
 For the general matrix, do it yourselves

41
 Input/output
 Input 123
456
>>> a = np.loadtxt(‘C:\\xyz\\infile.txt’)
>>> a
array([[1., 2., 3. ], If you don’t specify
[4. ,5. ,6. ]]) absolute paths, the
input/output files must
be in the current
 output working directory
>>> b = np.array([[3,2,1], [6,5,4]])
>>>savetxt(‘C:\\xyz\\outfile.txt’, b, ‘%d’) 321
654

%d: integer
%p.qf: float, with q digits after decimal
point, and keep at least p places

42
 Plotting
import numpy as np
import matplotlib.pyplot as plt

x = np.arange(0, 10, 0.2)

y = np.sin(x)
plt.plot(x, y)
plt.show()

1D array of values for variable

plt.plot(x, y)

1D array of values for function

43
 Plotting
 Make your figure more readable
import matplotlib.pyplot as plt
import numpy as np

def plotme():

x = np.linspace(0, 2, 100)

plt.plot(x, x, label='linear')
plt.plot(x, x**2, label='quadratic')
plt.plot(x, x**3, label='cubic')

plt.xlabel('x label')
plt.ylabel('y label')

plt.title("Simple Plot")

plt.legend()

plt.show()
44
 Programming Setting
 When you place Python program in a file, the best way to
operate on it is to use Python IDLE editor
 You can run the program directly from IDLE shell:
 Add a line to call a function
 Click run →run module
 You can also run any function you create in IDLE shell
 Suppose function abc(*) has been defined in file myfile.py
 Remove any call to abc(*) from myfile.py
 Insert the path ‘C://…/myfile.py’ in sys.path using
sys.path.insert(i, pth), which inserts path pth before index i
 Type import myfile on IDLE shell
 Type myfile.abc(*) to run function abc(*)
 If you want to use debugging tools to debug your
program, you must use the second method
45
 Programming Setting
 Always put
import numpy as np
before the main code of your .py file
 If you need to do plotting, place the following line
import matplotlib.pyplot as plt
 If you want to use functions defined in another file,
import it also

46