Lecture 16

MATLAB III: More Arrays

and Design Recipe
 Last Time (lectures 14 & 15)
Lecture 14: MATLAB I
• “Official” Supported Version in CS4: MATLAB 2018a
• How to start using MATLAB:
• CS Dept. Machines - run ‘cs4_matlab’
• Total Academic Handout (TAH) Local Install -
software.brown.edu
• MATLAB Online (currently 2019a) - matlab.mathworks.com
• Navigating the Workspace (command window, variables,
etc.)
• Data types in MATLAB (everything is a 64-bit double float
by default!)
• MATLAB Programs
• scripts (like Python)
• functions (file-based, outputs defined in signature)
• Anonymous functions and overwriting function names
(oops!)
2
 Last Time (lectures 14 & 15)
Lecture 15: MATLAB II
• Conditional Statements
• if...end
• if...else...end
• if...elseif...else...end
• switch...end
• Arrays and Matrices (default numeric type)
• scalars (1x1 value)
• 1D vectors (1xN or Nx1 arrays)
• 2D matrices (MxN)
• linspace(a, b, n) vs. first:step:max
• Array concatenation, slicing, and indexing
• Array Manipulation
• zero-padding
• removing elements
• row-to-column x(:)
• Size of arrays (numel and size; not length)
3
 Lecture 16 Goals: MATLAB III
• Multi-dimensional arrays:
• Applying built-in functions to matrices
• Scalar operations on matrices
• Element-wise operations on matrices
• Logical array comparisons
• Array indexing with ‘find’
• 3D arrays

4
 Arrays as function arguments

◻ Many MATLAB functions that work on single numbers will

also work on entire arrays; this is very powerful!
◻ Results have the same dimensions as the input, results are
produced “elementwise”
◻ For example:
>> av = abs([-3 0 5 1])
av =
3 0 5 1

5
 Powerful Array Functions

◻ There are a number of very useful function that are built-in

to perform operations on vectors, or column-wise on
matrices:
■ min the minimum value
■ max the maximum value
■ sum the sum of the elements
■ prod the product of the elements
■ cumprod cumulative product
■ cumsum cumulative sum

6
 min, max Examples

>> vec = [4 -2 5 11];

>> min(vec)
ans =
-2
>> mat = randi([1, 10], 2,4)
mat =
6 5 7 4
3 7 4 10
>> max(mat)
ans =
6 7 7 10

• Note: the result is a scalar when the argument is a vector; the result is a 1 x n
vector when the argument is an m x n matrix

7
 sum, cumsum vector Examples

◻ The sum function returns the sum of all elements; the

cumsum function shows the running sum as it iterates
through the elements (4, then 4+-2, then 4-2+5, and finally
4-2+5+11)
>> vec = [4 -2 5 11];
>> sum(vec)
ans =
18
>> cumsum(vec)
ans =
4 2 7 18

8
 What is the value of b?

a = [2 3 1; -2 0 -6; 8 7 -1];
b = min(a);

What is the value of b?

A) -6 B) [-2 0 -6]
C) [1 -6 -1] D) [-6 -6 -6]

9
 What is the value of b?

a = [2 3 1; -2 0 -6; 8 7 -1];
b = min(a);

What is the value of b?

A) -6 B) [-2 0 -6]
C) [1 -6 -1] D) [-6 -6 -6]

10
 What is the value of b?

a = [2 3 1; -2 0 -6; 8 7 -1];
b = min(a’);

What is the value of b?

A) -6 B) [-2 0 -6]
C) [1 -6 -1] D) [-6 -6 -6]

11
 What is the value of b?

a = [2 3 1; -2 0 -6; 8 7 -1];
b = min(a’);

What is the value of b?

A) -6 B) [-2 0 -6]
C) [1 -6 -1] D) [-6 -6 -6]

12
 What is the value of b?

a = [2 3 1; -2 0 -6; 8 7 -1];
b = min(a(:));

What is the value of b?

A) -6 B) [-2 0 -6]
C) [1 -6 -1] D) [-6 -6 -6]

13
 What is the value of b?

a = [2 3 1; -2 0 -6; 8 7 -1];
b = min(a(:));

What is the value of b?

A) -6 B) [-2 0 -6]
C) [1 -6 -1] D) [-6 -6 -6]

14
 sum, cumsum matrix Examples

◻ For matrices, most functions operate column-wise:

>> mat = randi([1, 10], 2,4)
mat =
1 10 1 4
9 8 3 7
>> sum(mat)
ans =
10 18 4 11
>> cumsum(mat)
ans =
1 10 1 4
10 18 4 11

The sum is the sum for each column; cumsum shows the
cumulative sums as it iterates through the rows

15
 prod, cumprod Examples

◻ These functions have the same format as

sum/cumsum, but calculate products
>> v = [2:4 10]
v =
2 3 4 10
>> cumprod(v)
ans =
2 6 24 240
>> mat = randi([1, 10], 2,4)
mat =
2 2 5 8
8 7 8 10
>> prod(mat)
ans =
16 14 40 80

16
 Overall functions on matrices

• When functions operate column-wise for matrices, make

nested calls to get the function result over all elements of a
matrix, e.g.:
>> mat = randi([1, 10], 2,4)
mat =
9 7 1 6
4 2 8 5
>> min(mat)
ans =
4 2 1 5
>> min(min(mat))
ans =
1

17
 Overall functions on arrays

◻ Alternatively, since linear indexing arranges all the

elements of an array into a column, you can also use this
approach.

>> m = max(A(:)) % Find max of A, regardless of

dim.

18
 Scalar operations

•Numerical operations can be performed on every element

in an array
•For example, Scalar multiplication: multiply every
element by a scalar
>> [4 0 11] * 3
ans =
12 0 33
•Another example: scalar addition; add a scalar to every
element
>> zeros(1,3) + 5
ans =
5 5 5

19
 Array Operations

• Array operations on two matrices A and B:

• these are applied between individual elements
• this means the arrays must have the same dimensions
• In MATLAB:
• matrix addition: A + B
• matrix subtraction: A – B or B – A
• For operations that are based on multiplication
(multiplication, division, and exponentiation), a dot must be
placed in front of the operator. Unless you’re doing linear
algebra, this point-wise approach is generally what you want.
• array multiplication: A .* B
• array division: A ./ B, A .\ B
• array exponentiation A .^ 2
• matrix multiplication: A*B is NOT an element-wise operation

20
 Logical Vectors and Indexing

• Using relational and logical operators on a vector or matrix

results in a logical vector or matrix
>> vec = [44 3 2 9 11 6];
>> logv = vec > 6
logv =
1 0 0 1 1 0
• Can use this to index into a vector or matrix, index and matrix
dimensions must agree (logical linear indexing also OK)
>> vec(logv)
ans =
44 9 11

21
 Element-wise logical operators

• | and & applied to arrays operate elementwise; i.e. go through

element-by-element and return logical 1 or 0
>> [1 2 3 -1 1]>[0 1 2 1 0]
ans = 1×5 logical array
1 1 1 0 1
• || and && are used for scalars

22
 True/False

◻ false equivalent to logical(0)

◻ true equivalent to logical(1)

◻ false(m,n) and true(m,n) create matrices of

all false or true values

23
 Logical Built-in Functions

◻ any, works column-wise, returns true for a column, if it

contains any true values
◻ all, works column-wise, returns true for a column, if all the
values in the column are true

>> M = randi([-5 100], m, n)

>> any(M<0 | M==5) % returns a 1 x n vector
% elements are true if corresponding
% column in M has any negative
% entries or any 5s in it.
>> all(M(:)>0) % true if all elements strictly positive

24
 Finding elements

◻ find finds locations and returns indices

>> vec
vec =
44 3 2 9 11 6
>> find(vec>6)
ans =
1 4 5

◻ find also works on higher dimensional arrays

[i,j] = find(M>0) % returns non-zero matrix

indices
ind = find(A>0) % returns linear array indices

25
 Comparing Arrays

◻ The isequal function compares two arrays, and returns

logical true if they are equal (all corresponding elements) or
false if not
>> v1 = 1:4;
>> v2 = [1 0 3 4];
>> isequal(v1,v2)
ans =
0
>> v1 == v2
ans =
1 0 1 1
>> all(v1 == v2)
ans =
0

26
 3D Matrices

◻ A three dimensional matrix has dimensions m x n x p

◻ Can create with built-in functions, e.g. the following
creates a 3 x 5 x 2 matrix of random integers; there are 2
layers, each of which is a 3 x 5 matrix
>> randi([0 50], 3,5,2)
ans(:,:,1) =
36 34 6 17 38
38 33 25 29 13
14 8 48 11 25
ans(:,:,2) =
35 27 13 41 17
45 7 42 12 10
48 7 12 47 12

27
 Functions diff and meshgrid

◻ diff returns the differences between consecutive elements in

a vector
◻ meshgrid receives as input arguments two vectors, and
returns as output arguments two matrices that specify
separately x and y values
>> [x y] = meshgrid(1:3,1:2)
x =
1 2 3
1 2 3
y =
1 1 1
2 2 2

Where could meshgrid be useful?

28
 Common Pitfalls

• Attempting to create a matrix that does not have the same number
of values in each row
• Confusing matrix multiplication and array multiplication. Array
operations, including multiplication, division, and exponentiation,
are performed term by term (so the arrays must have the same size);
the operators are .*, ./, .\, and .^.
• Attempting to use an array of double 1s and 0s to index into an
array (must be logical, instead)
• Attempting to use || or && with arrays. Always use | and & when
working with arrays; || and && are only used with logical scalars.

29
 Programming Style Guidelines

• Extending vectors or matrices is not very fast, avoid doing this too
much
• To be general, avoid assuming fixed dimensions for vectors , matrices
or arrays. Instead, use end and colon : in context, or use size and
numel

>> len = numel(vec);

>> [r, c] = size(mat);
>> last_col = mat(:, end);

• Use true instead of logical(1) and false instead of logical(0),

especially when creating vectors or matrices.

30
DESIGN Recipe

31
 Testing

• Even simple functions can be deceptively hard to verify as

correct just by “looking at them”
• However, it is easy to test functions on data you understand
(and know what the correct answer should be)
• As functions and programs (which may use lots of functions)
get more complicated this becomes very important

32
 assert

In MATLAB, the assert function allows one to easily perform a

test
assert(expr, message)
Stops execution and prints our the message when expr evaluates
to false.

33
 Examples

• test_triArea.m
• test_myQuadRoots.m

34
 Testing is Programming

• We’ve discovered developing tests first (before writing any

functions) often speeds the development process and helps
ensure programs work correctly
• In fact, designing tests should be viewed as a part of
programming even though you aren’t actively coding a
solution.

35
 Design Recipe

Design Recipe

1.Develop important Test Cases – (actually code them, requires

you to first create function header)
2.Code function body
3.Test!
4.Fix code, re-Test until working correctly

36
 Example: myFtoC

• Use the Design Recipe to solve the following problem:

“Write a function converts degrees Fahrenheit to degrees

Celsius.”

37
 Example: myFtoC

1. Write test_myFtoC
2. Write myFtoC
3. Run test_myFtoC
4. Fix code, re-test until working correctly
5. Look at code, identify any pertinent additional tests
6. Retest, until working correctly

Done!

38
 Example: myFtoC

◻ test_myFtoC.m
◻ myFtoC.m

39
 Example: quadMin

• Use the Design Recipe to solve the following problem:

“Write a function that finds x that minimizes

ax^2+bx+c
in the interval [L,R]. Assume a>=0, L<R.”

40
 Example: quadMin

“Write a function that finds x that minimizes

ax^2+bx+c
in the interval [L,R]. Assume a>=0, L<R.”

What kind of tests should we have?

What are the cases?

41
 Example: quadMin

• test_quadMin.m

42