Chapter 5: Vectorized Code
Exercises
1) The following code was written by somebody who does not know how to use
MATLAB efficiently. Rewrite this as a single statement that will accomplish exactly
the same thing for a matrix variable mat (e.g., vectorize this code):
[r c] = size(mat);
for i = 1:r
for j = 1:c
mat(i,j) = mat(i,j) * 2;
end
end
>> mat = mat * 2
2) Vectorize this code! Write one assignment statement that will accomplish exactly
the same thing as the given code (assume that the variable vec has been initialized):
result = 0;
for i = 1:length(vec)
result = result + vec(i);
end
>> result = sum(vec)
3) Vectorize this code! Write one assignment statement that will accomplish exactly
the same thing as the given code (assume that the variable vec has been initialized):
newv = zeros(size(vec));
myprod = 1;
for i = 1:length(vec)
myprod = myprod * vec(i);
newv(i) = myprod;
end
newv % Note: this is just to display the value
>> newv = cumprod(vec)
4) Create a 1 x 6 vector of random integers, each in the range from 1 to 20. Use
built‐in functions to find the minimum and maximum values in the vector. Also
create a vector of cumulative sums using cumsum.
Ch5Ex4.m
% Create a random vector and use built-in functions
% to find the minimum, maximum, and cumulative sums
vec = randi([1,20], 1,6)
min(vec)
max(vec)
cvec = cumsum(vec)
5) Write a relational expression for a vector variable that will verify that the last
value in a vector created by cumsum is the same as the result returned by sum.
>> vec = 2:3:17
vec =
2 5 8 11 14 17
>> cv = cumsum(vec)
cv =
2 7 15 26 40 57
>> sum(vec) == cv(end)
ans =
1
6) Create a vector of five random integers, each in the range from ‐10 to 10.
Perform each of the following using only vectorized code:
>> vec = randi([-10, 10], 1,5)
subtract 3 from each element
>> vec-3
count how many are positive
>> sum(vec > 0)
get the absolute value of each element
>> abs(vec)
find the maximum
>> max(vec)
7) Create a 3 x 5 matrix. Perform each of the following using only vectorized code:
>> mat = randi([-10 10], 3,5)
Find the maximum value in each column.
>> max(mat)
Find the maximum value in each row.
>> max(mat, [], 2)
>> max(mat')
Find the maximum value in the entire matrix.
>> max(max(mat))
8) Write a function called geomser which will receive values of r and n, and will
calculate and return the sum of the geometric series:
1 + r + r2 + r3 + r4 + ... + rn
The following examples of calls to this function illustrate what the result should be:
>> geomser(1,5)
ans =
6
>> disp(geomser(2,4))
31
geomser.m
function sgs = geomser(r, n)
% calculates the sum of the geometric series
% 1 + r + r^2 + r^3 + ... r^n
% Format of call: geomser(r, n)
% Returns sum of the series
v = 0:n;
terms = r .^ v;
sgs = sum(terms);
end
9) Generate a random integer n, create a vector of the integers 1 through n in steps
of 2, square them, and plot the squares.
Ch5Ex9.m
% Create a vector of integers 1:2:n where n is random
% square them and plot the squares
n = randi([1,50])
vec = 1:2:n;
vecsq = vec .^ 2;
plot(vecsq,'k*')
title('Squares of integers')
10) A vector v stores for several employees of the Green Fuel Cells Corporation their
hours worked one week followed for each by the hourly pay rate. For example, if
the variable stores
>> v
v =
33.0000 10.5000 40.0000 18.0000 20.0000 7.5000
that means the first employee worked 33 hours at $10.50 per hour, the second
worked 40 hours at $18 an hour, and so on. Write code that will separate this into
two vectors, one that stores the hours worked and another that stores the hourly
rates. Then, use the array multiplication operator to create a vector, storing in the
new vector the total pay for every employee.
>> hours = v(1:2:length(v))
hours =
33 40 20
>> payrate = v(2:2:length(v))
payrate =
10.5000 18.0000 7.5000
>> totpay = hours .* payrate
totpay =
346.5000 720.0000 150.0000
11) Write a function repvec that receives a vector and the number of times each
element is to be duplicated. The function should then return the resulting vector. Do
this problem using built‐in functions only. Here are some examples of calling the
function:
>> repvec(5:-1:1,2)
ans =
5 5 4 4 3 3 2 2 1 1
>> repvec([0 1 0],3)
ans =
0 0 0 1 1 1 0 0 0
repvec.m
function outvec = repvec(vec,n)
% Duplicates every number in a vector n times
% Format of call: repvec(vector, n)
% Returns a new vector
mat = repmat(vec,n,1);
outvec = reshape(mat,1,n*length(vec));
end
12) The mathematician Euler proved the following:
2 1 1 1
1
6 4 9 16
Rather than finding a mathematical proof for this, try to verify whether the
conjecture seems to be true or not. Note: there are two basic ways to approach this:
either choose a number of terms to add, or loop until the sum is close to 2/6.
Ch5Ex12.m
% Test Euler's theory that pi^2/6 = 1 + 1/4 + 1/9 + ...
fprintf('pi^2/6 is %.2f\n', pi^2/6)
% Try 25 terms
n = 25;
i = 1:n;
seriessum = sum(1./i.^2);
fprintf('The sum of %d terms is %.2f\n',n,seriessum)
When working with images, it is often necessary to “crop” one image to match the size
of another. Images are represented very simply in MATLAB as matrices of numbers.
However, these matrices are quite large. Depending on the resolution, the number of
rows and columns could easily be in the thousands. It is therefore extremely important
when working with image matrices to vectorize the code.
13) Write a script that will read from a file oldfile.dat into a matrix. It will create a
square matrix (same number of rows and columns) by deleting rows or columns as
necessary, and then write this new square matrix to a new file called squarefile.dat.
For example, if the original matrix is 4 x 6, the new matrix would be created by
deleting the fifth and sixth columns to result in a 4 x 4 matrix. Another example: if
the original matrix is 3 x 2, the third row would be deleted to result in a 2 x 2 matrix.
The script should be general and work regardless of the size of the original file and
should not use any loops or if statements. Create the data file first.
Ch5Ex13.m
% Reads a matrix from a file, creates a square matrix
% from it by deleting rows or columns as necessary and
% writes the resulting matrix to a new file
load oldfile.dat
mat = oldfile;
s = sort(size(mat));
newm = mat(1:s(1), 1:s(1));
save squarefile.dat newm -ascii
14) A file called “hightemp.dat” was created some time ago which stores, on every
line, a year followed by the high temperature at a specific site for each month of that
year. For example, the file might look like this:
89 42 49 55 72 63 68 77 82 76 67
90 45 50 56 59 62 68 75 77 75 66
91 44 43 60 60 60 65 69 74 70 70
etc.
As can be seen, only two digits were used for the year (which was common in the
last century). Write a script that will read this file into a matrix, create a new matrix
which stores the years correctly as 19xx, and then write this to a new file called
“y2ktemp.dat”. (Hint: add 1900 to the entire first column of the matrix.) Such a file,
for example, would look like this:
1989 42 49 55 72 63 68 77 82 76 67
1990 45 50 56 59 62 68 75 77 75 66
1991 44 43 60 60 60 65 69 74 70 70
etc.
Ch5Ex14.m
% Solve the Y2K problem!
load hightemp.dat
newmat = hightemp;
newmat(:,1) = newmat(:,1) + 1900;
save y2ktemp.dat newmat -ascii
15) Write a script that will prompt the user for a quiz grade and error‐check until
the user enters a valid quiz grade. The script will then echo print the grade. For this
case, valid grades are in the range from 0 to 10 in steps of 0.5. Do this by creating a
vector of valid grades and then use any or all in the condition in the while loop.
Ch5Ex15.m
% Prompt the user for a quiz grade and error-check
% until a valid grade is entered; using any or all
quiz = input('Enter a quiz grade: ');
validgrades = 0: 0.5: 10;
while ~any(quiz == validgrades)
quiz = input('Invalid! Enter a quiz grade: ');
end
fprintf('The grade is %.1f\n', quiz)
16) Which is faster: using false or using logical(0) to preallocate a matrix to all
logical zeros? Write a script to test this.
% Test to see whether logical(0) or false is faster
fprintf('Start with logical(0)\n\n')
tic
mat(1:100, 1:100) = logical(0);
toc
fprintf('Now for false \n\n')
tic
mat2(1:100, 1:100) = false;
toc
17) Which is faster: using a switch statement or using a nested ifelse? Write a
script to test this.
Ch5Ex17.m
% Test which is faster: switch or nested if-else
fprintf('First for if-else\n\n')
tic
for i = 1:1000
for j = 1:4
if j == 1
result = 11;
elseif j == 2
result = 22;
elseif j == 3
result = 46;
else
result = 6;
end
end
end
toc
fprintf('Now switch\n\n')
tic
for j = 1:1000
for k = 1:4
switch k
case 1
res = 11;
case 2
res = 22;
case 3
res = 46;
otherwise
res = 6;
end
end
end
toc
18) Vectorize this code:
n = 3;
x = zeros(n);
y = x;
for i = 1:n
x(:,i) = i;
y(i,:) = i;
end
>> [x y] = meshgrid(1:3,1:3)
19) A company is calibrating some measuring instrumentation and has measured
the radius and height of one cylinder 10 separate times; they are in vector variables
r and h. Use vectorized code to find the volume from each trial, which is given by
Πr2h. Also use logical indexing first to make sure that all measurements were valid
(> 0).
>> r = [5.501 5.5 5.499 5.498 5.5 5.5 5.52 5.51 5.5 5.48];
>> h = [11.11 11.1 11.1 11.12 11.09 11.11 11.11 11.1 11.08 11.11];
>> all(r>0 & h>0)
ans =
1
>> vol = pi * r.^2 .* h
