mattoneal.com

One way to speed up the executing times of your coding projects is to use MATLAB and Octave’s native vectorization capabilities. In short, this is one of the significant advantages of using a vector- or matrix-based programming language over the more popular object oriented languages. What do I mean by vectorization? I’ll show you with a quick example.

Maybe you need to sum all the numbers from 1 to 1,000,000. A simple for loop will do the trick.

clear
tic();
 
x = 0;
 
for n = 1:1e6
	x = x + n;
end
x
exec_time = toc()

On my computer the above took 1.249 seconds to execute. Not a tremendous amount of time, but how can we make it faster? By not using a loop, and operating directly on vectors. Try the below…

clear
tic();
 
s = sum(1:1e6);
 
exec_time = toc()

This code executed in 0.0192 seconds, which is more than 50 times faster. The difference between a second and 0.019 seconds may not seem like much, but what if you had to run something like the above within another loop for a hundred or more iterations? Now the difference becomes code that could execute in about 3 seconds compared to 3 minutes. The more complex your coding gets, the more important it is to try to use MATLAB and Octave’s inherent strengths whenever possible.

Now, it may be impossible to eliminate loops from your coding all together, but ask yourself whenever you’re writing a for or while loop, are you doing it to create another vector? If you’re using it with an embedded conditional (IF/ELSE), chances are you may be stuck with it. But if you are looping through a vector to create another vector, chances are you could speed it up by performing the operation directly on the initial vector.

And so it goes. Happy coding!

Often you’ll have a need for recording the data your programs generate. MATLAB and Octave have a few ways to do this. We can best show this with a quick example.

The following code will generate a matrix 10,000 by 6. On my current hardware this takes about 2 seconds to execute.

% Sample code to generate a large matrix
clear 
 
for i = 1:10000
	a = sin(i*pi/180);
	b = cos(i*pi/180);
	c = tan(i*pi/180);
	temparray(i, 1:6) = [a, b, c, a^2, b^2, c^2];
end

In MATLAB and Octave, the process of outputting data to files is called “exporting.” Reading data in is likewise called “importing.”

The easiest way to save the above output to a file is to use the save function and add it directly to the code. Adding the following line just below the for loop would do the trick.

save test.txt temparray -ascii

The format is: save filename variable -format. The above will result in a text file called test.txt which will contain the variable “temparray.” You can open or view this in any text editor. The default delimiter is a space but if you’d like to save it as tab-delimited data, you would simply change the line to the following:

save test.txt temparray -ascii -tabs

If you do not specify the variable, all variables in the current workspace will be saved to the file. The above could also be executed from the command line, as long as you had recently run the code and the variable “temparray” was still assigned the value you need. Worth noting is that within Octave or MATLAB, if you save without using the -ascii tag, the default will be to save the data in a proprietary binary format and you will have something like the below text at the top of your document:

# Created by Octave 3.2.4, Thu Mar 10 00:24:15 2011 EST
# name: temparray
# type: matrix
# rows: 10 000
# columns: 6
1.7452e-02   9.9985e-01   1.7455e-02   3.0459e-04   9.9970e-01   3.0468e-04
3.4899e-02   9.9939e-01   3.4921e-02   1.2180e-03   9.9878e-01   1.2195e-03
5.2336e-02   9.9863e-01   5.2408e-02   2.7391e-03   9.9726e-01   2.7466e-03
...

This may be inconvenient for your purposes (for example, if you are using the data with an external program), but if you are going to use the “load” command (described later) to import this file, it doesn’t matter that this header information is present. If it is a hindrance, you can always save the data using the -ascii tag.

Other ways to export data to a file is with the fwrite, csvwrite and dlmwrite commands. Both programs have detailed help files on these commands but in short, they are for appending to files (fwrite), saving as comma-separated-values (csvwrite) and using specific delimiters (dlmwrite).