Data Entry in Matlab

• Explicitly: Typing "A = [11 12 13; 21 22 23]" gives a 2x3 matrix, which can also be entered using:

  >> A = [  11  12  13
            21  22  23 ]
  

  Commas or blank spaces separate row entries inside the matrix-builder brackets "[]", while semicolons or carriage returns separate the column entries. So as shown earlier, x = [1 2 3] is a row vector, while x = [1; 2; 3] is a column vector.
• By concatenation. If V is an nxn array and x is an nx1 vector, then

  A = [V,x]
  

  is an nx(n+1) array, and

  B = [x' ; V] 
  

  is an (n+1)xn array. It is real common to build up a vector or array step-by-step inside a loop using matrix builder brackets.
• From an ascii file (created with an editor or as output of another program). You can use format statements to write out characters like "A = [" just before outputting the data and a "];" right afterwards. Name the output file something like indata.m, enter Matlab, and type "indata" at the prompt. That will set up an array A containing the data. Here is an example of how this works. Suppose that I have written a C program that produces timings of an operation, one per line. If the data looks like

     1.23132E+01
     8.54579E+00
     1.21672E+01
     .
     .
     .
     8.53653E+00
     8.56410E+00
  

  then I use my trusty text editor to insert lines before and after the text in the output file:

  A = [
     1.23132E+01
     8.54579E+00
     1.21672E+01
     .
     .
     .
     8.53653E+00
     8.56410E+00
  ];
  
  subplot(2,1,1);
  
  plot((1:length(A)),A, '*')
  title('Times for Small Memory Job')
  ylabel('Seconds')
  xlabel('Run Number')
  
  subplot(2,1,2);
  hist(A,40)
  title('Times for Small Memory Job')
  xlabel('Seconds')
  ylabel('Number of Runs')
  

  This creates a plot of those times, and a histogram:

  Plot of times from sample run

  

  Another way doesn't involve modifying the file indata.m. At the Matlab prompt, enter

  load indata.m
  

  That will create an array with the name "indata". Beware that this is the opposite invocation you normally would do; the load is done on indata.m, not with the suffix .m removed.

• From Matlab matrix generation statements: rand, randn, ones, zeros, hilbert, .... These generate matrices with characteristics you would expect. Don't know what ones() generates? Use the help command!
• From calls to Fortran or C programs. Matlab can directly invoke those languages, with results returned to Matlab. More on that later.
• If data was created within Matlab, can use the save command:

  save filename var1 var2 ... varn
  

  saves the variables listed in a binary file called "filename.mat".
  • If variable list is omitted, all of workspace is saved
  • Can move the binary file filename.mat from machine to machine.
  • If no filename is given, save in "matlab.mat"
  You can restore the data using

  load filename x1 x2 ... xn
  

  with the same kind of restrictions and uses as the save command.

As an important reminder ...
Beware: By default, the result of all assignment operations are printed. Appending a semicolon (;) to end of an assignment suppresses this printing. The command sequence

>> A = zeros(100,100);
>> for j = 1:100
>>    for i = 1:100
>>       A(i,j) = 1.0/(i+j-1)
>>    end
>> end

A better mechanism is to use the save and load commands. By default they will save/load a binary file containing all of the variables in your current workspace - sort of a snapshot of the session. You can also save just specified variables, save to an ascii format file, etc. A good use for save is when you hit a bug while working, and cannot get around it. Save your workspace along with a note of what you were trying to do, so that it can be easily reloaded and demonstrated to me (or to Mathworks, if it is their bug!)

• Next page: Matlab Array Notation
• Class home page: A321