\documentclass[autumn,nocolorBG]{prosper}
\title{The Briefest of Introductions to MATLAB}
\subtitle{A Computational Environment for the Behavioral Sciences}
\author{Lawrence Hubert}
\institution{University of Illinois\\ These slides are available at: \\ cda.psych.uiuc.edu/multivariate\_fall\_2013}
\begin{document}
\maketitle
\begin{slide}{MATLAB (Matrix Laboratory)}
This is the computational environment that is becoming the universal standard for the Behavioral Sciences. It is already the standard for all the Engineering and Computer Science disciplines.
\bigskip
Developed and maintained by the MathWorks, physically located in Natick, Massachusetts (just outside of Boston).
\bigskip
One of the least predatory companies I've ever encountered, with
just great pricing for students and academic settings generally.
This is in contrast to SPSS, SAS, and the like, or for that matter,
Adobe.
\bigskip
All the help, manuals, and so on you will ever need (and for free), are at \verb+http://www.mathworks.com+
\end{slide}
\begin{slide}{}
Also, see
\smallskip
\verb+www.mathtools.net+
\smallskip
This is a site maintained by the MathWorks for computational things generally.
\bigskip
In comparison to closed software systems like SPSS, SAS, and SYSTAT (where you don't get to see or modify the actual code, or chose different options if they don't deem it necessary), the very (old) cliched proverb to follow is appropriate (edited for the original sexist language):
\bigskip
``Give a person a fish; you will have fed him/her for today. Teach a person to fish, and you have fed him/her for a lifetime.''
\end{slide}
\begin{slide}{}
With some facility with Matlab, you are no longer dependent on what (obsolete) fish someone else tries to feed you, e.g., ``in $K$-means clustering, local optima are not an issue, so we will give you only one solution'' WRONG!
\bigskip
You can ``roll your own'' analyses ---
\smallskip
Extra credit: what generation does this phrase come from?
\bigskip
More indirectly, the MathWorks gives generously to public radio in Boston (WBUR) ---
\bigskip
They may be the ``Ben and Jerry's'' of software environments ---
\end{slide}
\begin{slide}{}
There is an enormous amount of detail that we could go over, but I will mention or introduce only what I have used and found valuable.
\bigskip
The various demos we do now are only intended for the mechanics; you don't need to understand all the nuances of the things introduced.
\bigskip
What might be a good place to start is the free Getting Started Manual (\verb+matlab_primer.pdf+, available where these notes are placed).
\bigskip
But if you stick with me for the next few hours, you should have
enough to at least begin learning ``on the job'', using all the
``help'' material that is freely available.
\end{slide}
\begin{slide}{Basic MATLAB}
\begin{itemize}
\item The Mother of all Calculators ---
\medskip
where everything can be done with matrices (in fact, the basic unit of analysis is a matrix).
\medskip
This is where we start, using a simple example to show MATLAB as a powerful computational scratchpad. For our carry-along illustration, the same data set from Psychology 406 is used that introduced multiple regression (at least when I taught 406).
\medskip
There are many built-in functions, including lots of matrix functions, to get our work done (MATLAB started as an interface to LINPACK and EISPACK, both linear algebra packages, and continues in this tradition today using LAPACK and BLAS).
\end{itemize}
\end{slide}
\begin{slide}{}
\begin{itemize}
\item Programming language ---
\smallskip
it is very straightforward to use, but extremely powerful and much easier than, for example, C or Fortran. (There is no explicit array dimensioning needed, or to indicate and keep track of whether you are dealing with real or integer numbers.)
\medskip
There is \emph{nothing} you can do in Fortran or C that you can't do in MATLAB (and in a much easier fashion).
\medskip
It is an interpreted language (so it translates and executes the statements ``on the fly''); this is in contrast to compiled C or Fortran (or Basic or Pascal, among others), which produce stand-alone executables.
\end{itemize}
\end{slide}
\begin{slide}{}
Given recent enhancements, interpreted MATLAB \emph{is as fast as} compiled C.
\medskip
Generally, we can do better speedwise with compiled Fortran (about a tenfold increase). Also, we can now do so with automatic conversions from MATLAB code to Fortran (the old \verb+.dll+'s). I will mention how in a moment.
\medskip
There is one overriding advantage of writing your own functions in
MATLAB --- they are easily modifiable \verb+ascii+ files. So, if
someone wishes to run these, I just send them over email, and he/she
can just ``fire-up'' MATLAB (a non-hobbled student version with
several toolboxes [e.g., Statistics and Optimization] costs about a
\$100) ---
\end{slide}
\begin{slide}{}
or use one of the ``free'' alternatives that will run M-files --- OCTAVE, for example:
\bigskip
\verb+http://www.octave.org+
\bigskip
I've been a Fortran programmer for over four decades, and that is long enough to keep track of matrix dimension sizes or to worry about what is an integer and what is not. I can't tell you emphatically enough how much easier it is now to write my code in MATLAB.
\bigskip
If I need the speed, I can translate to callable Fortran subroutines. But usually, you can just stay with the MATLAB code (written in the form of what are called M-files) because that is generally fast enough.
\end{slide}
\begin{slide}{}
In Fortran, I was a ``spaghetti coder'', with lots of ``go to''s (with statement numbers). In fact, I was once known as the ``go to'' kind-of guy --- (this is a joke)
\bigskip
In MATLAB, there are no ``go to''s and no statement numbers. When you use the (chroma-coded) Editor, structured programming is done automatically for you, with nice indentations and things lined up appropriately for easy reading.
\bigskip
I would gladly be a poster child for the MathWorks; they could put my picture on a milk carton with the caption: ``Have you seen this Fortran programmer?''
\end{slide}
\begin{slide}{}
\begin{itemize}
\item Provides great graphics ---
\bigskip
2- and 3-D plotting, visualization, animation, image processing.
\smallskip
In 2-D plotting of prime importance in presenting data and statistical results, graphs can be edited easily with tools directly from the MATLAB figure window and exported to common file types --- \verb+.eps+, \verb+.jpg+, and \verb+.pdf+, for example.
\smallskip
Files of these types can be embedded into Word documents (if you must), or better, using \LaTeX{}.
\smallskip
The MATLAB figure format file type (called \verb+.fig+) can be used so the figure can be reedited in MATLAB
\end{itemize}
\end{slide}
\begin{slide}{}
\begin{itemize}
\item Graphical User Interface Development Environment (GUIDE) ---
\bigskip
You can develop your own data collection GUIs or demonstrations. A prime example of this would be incorporating the (free, and freely available, open-source) (visual) Psychophysics Toolbox (primarily developed by David Brainard, now at U. Penn., Psychology):
\bigskip
\verb+http://psychtoolbox.org+
\bigskip
For more, take the course in MATLAB programming from our own Alejandro Lleras.
\end{itemize}
\end{slide}
\begin{slide}{}
\begin{itemize}
\item Application Program Interface (API) ---
\smallskip
for calling C, Fortran, Java routines within the MATLAB environment.
\medskip
We mentioned this earlier in the context of speeding up MATLAB code
using Fortran produced \verb+.dll+s (actually, they now go by the
extension \verb+.mexw32+ on windows machines using the current
version of MATLAB [this is release 2013a]).
\medskip
The conversion program (in the way of M-files) is called \verb+matlab2fmex.m+, and written by Benjamin Barrowes:
\smallskip
\verb+sourceforge.net/projects/matlab2fmex+
\end{itemize}
\end{slide}
\begin{slide}{Toolboxes}
A Toolbox is a collection of M-files that do analysis within some specific area. Typically, the source code for the M-files is available for modification, even when the original toolbox may have been purchased (for real money even, as Yogi Berra says).
\bigskip
The commercial Toolboxes through the MathWorks are all on our site license. The first five Toolboxes I list below are probably the most germane for us (for example, I have paid for a separate license for these for years so they can easily be put on my laptop without worry about access to the license server).
\bigskip
Statistics; Optimization; Image Processing; Curve Fitting; Bioinformatics; Global Optimation
\end{slide}
\begin{slide}{}
Others of interest ---
Neural Networks; Splines; Wavelets, Symbolic Math; Mapping (Geographical); Parallel Computing; Econometrics
\bigskip
Also note ---
Matlab Coder (this converts M-files directly to C or C++ code, and generates stand-alone applications that can be distributed)
\end{slide}
\begin{slide}{}
Free (open source) Toolboxes ---
\bigskip
\verb+www.mathworks.com/matlabcentral+ --- several thousand M-files contributed by users.
\smallskip
\verb+www.mathtools.net/MATLAB+
\bigskip
Auditory Perception; ILAB (eye movement analyses); Brainstorm
(MEG/EEG); Chemometrics; N-way Toolbox; Smoothing; Econometrics;
Spatial Statistics; NURBS; ICA (EEG/ERP -- independent component
analysis); LYNGBY (fMRI analyses); NIH CORTEX; NMRLAB;
Psychophysics; Speech Processing; EMEGS (for EEG and MEG); SPM8
(Statistical Parametric Mapping -- neuroimaging); fmristat; surfstat
\end{slide}
\begin{slide}{My Toolboxes}
I have put a lot of material (we will talk about) at my web site:
\smallskip
\verb+cda.psych.uiuc.edu/multivariate_fall_2013+
\smallskip
`cda' stands for ``combinatorial data analysis''
\bigskip
One published item: Hubert, L., Arabie, P., \& Meulman, J. (2006). \emph{The structural representation of proximity matrices with MATLAB}. SIAM: Philadelphia.
\bigskip
An earlier draft: (directory: mono\_stuff\_11\_1\_05)
\verb+siam_final_submission_r1.pdf+
\smallskip
The M-files directory: \verb+srpm_mfiles+
\end{slide}
\begin{slide}{}
Two other Toolboxes ---
\bigskip
Cluster Analysis:
\verb+5283-Millsap-Ch20.pdf+
\verb+clusteranalysis_mfiles+
\bigskip
Unidimensional Scaling:
\verb+new_unidimensionalscaling_chapter.pdf+
\verb+new_unidimensionscaling_mfiles+
\bigskip
These will be handed out later in the semester.
\end{slide}
\begin{slide}{}
There are three books that are relevant to statistical issues and MATLAB; all three have downloadable M-files to go along with these sources.
\medskip
\emph{Computational Statistics Handbook with MATLAB (Second
Edition)} Wendy L. Martinez/Angel R. Martinez
Chapman \& Hall/CRC, 2008
\medskip
\emph{Exploratory Data Analysis with MATLAB} (Second Edition) Wendy L. Martinez/Angel R. Martinez
Chapman \& Hall/CRC, 2010
\medskip
\emph{Multi-way Analysis: Applications in the Chemical Sciences} Age Smilde/Rasmus Bro/Paul Geladi
Wiley, 2004
\end{slide}
\begin{slide}{}
There are several other very good sources for learning about things MATLAB, for free --
\bigskip
The large collections of demos that come with MATLAB and the Toolboxes
\bigskip
The great number of Webinars that you can see for free, covering all aspects of MATLAB and the various toolboxes. We list below a few of the possibilities:
\smallskip
\emph{Introduction to MATLAB}
\emph{Data Analysis with Curve Fitting and Statistics Toolboxes}
\emph{Applied Optimization Using MATLAB}
\end{slide}
\begin{slide}{}
The easiest way to get help ``on the fly'' is to issue the command \verb+help+ in the command window; a collection of help directories come up.
\bigskip
The ones I use most often are:
\verb+help matfun+ --- all the basic matrix functions
\verb+help datafun+ --- all the basic data analysis functions
\verb+help graph2d+ --- all the basic 2-dimensional plotting functions
\verb+help stats+ --- all the functions in the MATLAB Statistics Toolbox
\verb+help optim+ --- all the functions in the MATLAB Optimization Toolbox
\end{slide}
\begin{slide}{MATLAB as a Calculator}
Some of the capability of MATLAB as a calculator will be illustrated with the same data I use in Psychology 406 to introduce simple correlation and regression, least-squares, the general linear model, and all of the related concepts.
\bigskip
I have put these data into a file (\verb+ascii+) with Notepad, with the name: \verb+community_data.dat+
\bigskip
Note: all the personal files that you use or create in MATLAB should be on, say, your own flash drive; you can then make this your current directory (for the calculator use of MATLAB; creating your own M-files to use, and so on).
\end{slide}
\begin{slide}{}
\begin{tabular}{cccc}
community & accidents(100s) & vehicles(1000s) & police \\ \hline
1 & 1 & 4 & 20 \\
2 & 4 & 10 & 6 \\
3 & 5 & 15 & 2 \\
4 & 4 & 12 & 8 \\
5 & 3 & 8 & 9 \\
6 & 4 & 16 & 8 \\
7 & 2 & 5 & 12 \\
8 & 1 & 7 & 15 \\
9 & 4 & 9 & 10 \\
10 & 2 & 10 & 10 \\
\end{tabular}
\end{slide}
\begin{slide}{}
The \verb+.dat+ extension is important (so turn on ``show known file types'' in Windows --- tools/Folder Options/view --- and uncheck the box next to ``Hide extensions for known file types''). The extension tells MATLAB it is an \verb+ascii+ file and can be loaded with the command: \verb+load community_data.dat+
\bigskip
Once done, a $10 \times 3$ matrix called \verb+community_data+ is then available in your workspace to operate on and use.
\bigskip
The verbatim contents of \verb+community_data.dat+ are:
\end{slide}
\begin{slide}{}
\begin{verbatim}
1 4 20
4 10 6
5 15 2
4 12 8
3 8 9
4 16 8
2 5 12
1 7 15
4 9 10
2 10 10
\end{verbatim}
\end{slide}
\begin{slide}{The Desktop}
When the default Desktop is called up, we have the:
\smallskip
command window: this is where everything happens and all commands are entered
\bigskip
command history: if you would care to, you could reuse previous commands completed earlier that are saved here
\bigskip
workspace: this is where all the variables (matrices) exist that you can work on
\bigskip
current directory: set this to your current work area. I would suggest that this be a flash drive for all your ongoing stuff (with, obviously, continual backups to other places, right!)
\end{slide}
\begin{slide}{}
You can access all the help possibilities you will ever need from the Desktop.
\bigskip
We will now load up MATLAB and show you the desktop. We will continue to flip back and forth between MATLAB and these slides, so have the hard copies in front of you.
\end{slide}
\begin{slide}{Basic Data Analysis}
The elementary (built-in) data analysis functions generally operate in a column-wise orientation on a matrix:
\begin{verbatim}
load community_data.dat
community_data
mean(community_data)
\end{verbatim}
\bigskip
The functions \verb+median+, \verb+mode+, \verb+max+, \verb+min+, \verb+std+, and \verb+var+, work the same way.
\bigskip
To do it row-wise, use the transpose operation `` $\prime$ ''; for example, \verb+var(community_data')+
\end{slide}
\begin{slide}{}
The following four executed statements illustrate a few ways matrices operate with MATLAB that are relevant to us.
\medskip
the first shows how matrices may be entered directly by hand;
\medskip
the second is the construction of a matrix by concatenation;
\medskip
the third illustrates that scalar multiplication of a matrix is extended by MATLAB to scalar addition;
\medskip
the fourth illustrates that matrices are indexed in a row-column order.
\end{slide}
\begin{slide}{}
\tiny
\begin{verbatim}
>> A = [1 2 3;2 1 3;3 1 2]
A =
1 2 3
2 1 3
3 1 2
>> B = [A 3*A;A-(2*A) A+A]
B =
1 2 3 3 6 9
2 1 3 6 3 9
3 1 2 9 3 6
-1 -2 -3 2 4 6
-2 -1 -3 4 2 6
-3 -1 -2 6 2 4
\end{verbatim}
\end{slide}
\begin{slide}{}
\tiny
\begin{verbatim}
>> C = A + 10
C =
11 12 13
12 11 13
13 11 12
>> A(2,3)
ans =
3
\end{verbatim}
\normalsize
\end{slide}
\begin{slide}{}
The two functions \verb+cov+ and \verb+corrcoef+ give the ($3 \times 3$) covariance and correlation matrices.
\bigskip
\begin{verbatim}
corr_community_data = ...
corrcoef(community_data)
cov_community_data = ...
cov(community_data)
correlation_accident_vs_vehicle = ...
corr_community_data(1,2)
cov_accident_vs_police = ...
cov_community_data(1,3)
\end{verbatim}
\end{slide}
\begin{slide}{}
Let's see if we can get to the solution of the normal equations:
\[ \mathbf{b} = (\mathbf{X}'\mathbf{X})^{-1}\mathbf{X}\mathbf{Y} \]
We wish to predict the dependent variable, number of accidents (placed in $\mathbf{Y}$, $10 \times 1$), from the number of vehicles and the number of police (second and third columns of the $10 \times 3$ design matrix $\mathbf{X}$, with the first column containing all ones for the additive constant). The $3 \times 1$ regression vector $\mathbf{b}$ contains, in order, the additive constant and the regression coefficients on the number of vehicles and number of police.
\end{slide}
\begin{slide}{}
\begin{verbatim}
Y = community_data(:,1)
X = [ones(10,1),community_data(:,2),...
community_data(:,3)]
b = (inv(X'*X))*(X'*Y)
pred_values = X*b
correlations = ...
corrcoef([Y,pred_values])
R_squared = (correlations(1,2))^2
\end{verbatim}
\end{slide}
\begin{slide}{Other Matrix Operations}
\begin{verbatim}
det(X'*X)
rank(community_data)
trace(X'*X)
inv(X'*X)
\end{verbatim}
\bigskip
Some special matrices:
\begin{verbatim}
zeros(10,10) [ones(10,10)]
rand(10,10) [randn(10,10)]
eye(10)
\end{verbatim}
\end{slide}
\begin{slide}{For Multivariate Analysis}
$ [\mathbf{V},\mathbf{D}] = \mathrm{eig}(\mathbf{A})$, where $\mathbf{A}$ = $\mathbf{V}\mathbf{D}\mathbf{V}'$, for $\mathbf{A}$ square; $\mathbf{V}$ is orthogonal and contains eigenvectors (as columns); $\mathbf{D}$ is diagonal and contains the eigenvalues.
\bigskip
$ [\mathbf{U},\mathbf{S},\mathbf{V}] = \mathrm{svd}(\mathbf{B})$, where $\mathbf{B}$ = $\mathbf{U}\mathbf{S}\mathbf{V}'$; the columns of $\mathbf{U}$ and the rows of $\mathbf{V}'$ are orthonormal; $\mathbf{S}$ is diagonal and contains the non-negative singular values (ordered from \emph{largest to smallest}).
\end{slide}
\begin{slide}{Some Details}
The semicolon (;) at the end of a statement, suppresses output.
\bigskip
MATLAB is case-sensitive.
\bigskip
The up-arrow key ($\uparrow$) retrieves the previous command (for editing purposes, possibly).
\bigskip
\verb+clc+ clears the screen.
\bigskip
An ellipsis ($\ldots$) allows a long command to continue on the next line.
\end{slide}
\begin{slide}{}
The colon operator (:) does a lot of things:
\smallskip
[1:10] produces 1 2 3 4 5 6 7 8 9 10 ---
\smallskip
A(:,1) selects all rows and just the first column of A.
\smallskip
gives index ranges in ``for'' loops: \verb+for i = 1:n+
\bigskip
NaN stands for ``not a number'', and you get it, for example, from 0/0 ---
\bigskip
There are many ways of getting a matrix into the workspace. We have mentioned two: the \verb+ascii+ format in a \verb+.dat+ file and using the \verb+load+ command; or enter directly as a matrix, e.g.,: \verb+[2 2 2;3 4 2;1 6 7]+ with semicolons separating rows, and blanks or commas separating row entries.
\end{slide}
\begin{slide}{}
There is also an import wizard for \verb+.xls+ files (among others).
\bigskip
Generally, we will construct matrices through various operations or concatenations:
\smallskip
for rows: \verb+[ --- ; --- ; --- ]+
\smallskip
for columns: \verb+[ --- , --- , --- ]+
\bigskip
You can save the workspace, or part of it, as a \verb+.mat+ file, and reload it later. Or you can send it to someone else, and they could reload into their MATLAB environment.
\end{slide}
\begin{slide}{}
\verb+.mat+ files are binary so they are not easily readable.
\bigskip
There is also a \verb+diary+ function that can save your output to a file (when \verb+diary+ is \verb+on+). When \verb+diary+ is \verb+off+, the output is not saved to a file. Use \verb+diary filename+; if no file is specified, the file is just called \verb+diary+; use \verb+diary off+ to stop saving the output.
\bigskip
Don't be afraid to use (a lot of redundant) parentheses to clarify the order of operations; it will never hurt.
\bigskip
Never, never, ever, rely on the order of operation precedence to clarify an ambiguous statement --- make it explicit with parentheses. Trust me on this one; you will mess up big-time, many time(s), if you don't.
\end{slide}
\begin{slide}{MATLAB for Plots}
One very powerful feature of MATLAB is its (2 -- and 3 -- dimensional) plotting capabilities, along with the very nice environment to edit and label these plots extensively, and eventually, to save plots in a variety of file formats: \verb+.eps+. \verb+.pdf+, \verb+.jpg+, among others.
\bigskip
I would suggest using \verb+.eps+ always --- it is the (or at
least, my) preferred way to put graphics into \TeX{} and \LaTeX{}
documents; it is also editable and savable in all the usual drawing
programs (e.g., Illustrator).
\end{slide}
\begin{slide}{}
\begin{verbatim}
plot(community_data(:,2), ...
community_data(:,1),'ko')
for i = 1:10
objectlabels{i,1} = int2str(i);
end
text(community_data(:,2), ...
community_data(:,1), ...
objectlabels,'fontsize',10, ...
'verticalalignment','bottom')
\end{verbatim}
\end{slide}
\begin{slide}{}
\begin{verbatim}
ylabel('number of accidents')
xlabel('number of vehicles')
title('Scatterplot of Accidents ...
versus Vehicles')
axis([0,20,0,6])
text(14,1,'{\it Note the italics}')
gtext('{\bf Note the boldface}')
\end{verbatim}
\end{slide}
\begin{slide}{Some Plotting Points}
There are many tools and ways for editing your plots, both from the
command line and by using the simple Adobe Illustrator-like
interface in the Figure window. Also, on exporting the file, you can
specify the size.
\smallskip
If you stay in one of the formats Illustrator can edit (e.g.,
\verb+.eps+), you can use Illustrator to edit your graphics as will.
\smallskip
Once you have a plot the way you like it, you could even generate the M-file that produced it --- which would then be ready to accept another set of data.
\smallskip
\TeX{} and \LaTeX{} notation can be used in the plots (for symbols, italics, formulas, and so on). Text can be placed at certain coordinates with \verb+text(x,y,'string')+, or with cross-hairs with \verb+gtext('string')+.
\end{slide}
\begin{slide}{Script M-files}
A script M-file is a collection of statements and commands placed in an file that can be executed when the file name is typed in the command window.
\bigskip
Scripts are easily reusable (and changeable). They operate on and produce variables in the workspace.
\bigskip
As an example, we use \verb+ganmaa_script.m+, which produces
boxplots for some two-independent sample data I was helping someone
with.
\bigskip
Make sure you name your scripts meaningfully, so you will know what they are the next time you look into your current directory (not, e.g., \verb+script_one.m+).
\end{slide}
\begin{slide}{}
\begin{verbatim}
load pre_vitd_level.dat
load post_vitd_level.dat
pre_vitd_level(:,1)
post_vitd_level(:,1)
boxplot([pre_vitd_level(:,1), ...
post_vitd_level(:,1)],'notch', ...
'on','labels',{'Before','After'})
title('Vitamin D Levels in Children ...
Before and After Drinking Vitamin D ...
Fortified Whole Milk')
ylabel('25(OH)D, ng/ml')
\end{verbatim}
\end{slide}
\begin{slide}{Function M-files}
A function M-file takes input argument(s) and produces output arguments. The form of two functions, \verb+targlin.m+ and \verb+proxrand_altcomp.m+, shows the general pattern.
\medskip
\verb+targlin.m+ produces a symmetric proximity matrix of size $n \times n$ containing distances between equally and unit-spaced positions along a line;
\medskip
\verb+proxrand_altcomp.m+ produces a symmetric proximity matrix with entries that are a random permutation of those in a symmetric input matrix.
\medskip
If you type, for example, \verb+help targlin+, you will get the help header comments echoed. These are the lines ``commented out'' with ``\%''.
\end{slide}
\begin{slide}{targlin.m}
\tiny
\begin{verbatim}
function [targlinear] = targlin(n)
% TARGLIN produces a symmetric proximity matrix of size
% $n \times n$, containing distances
% between equally and unit-spaced positions
% along a line: targlinear(i,j) = abs(i-j).
%
% syntax: [targlinear] = targlin(n)
targlinear = zeros(n,n);
for i = 1:n-1
for j = (i+1):n
targlinear(i,j) = abs(i-j);
targlinear(j,i) = targlinear(i,j);
end
end
\end{verbatim}
\end{slide}
\begin{slide}{proxrand\_altcomp.m}
\tiny
\begin{verbatim}
function [randprox] = proxrand_altcomp(prox)
% PROXRAND_ALTCOMP produces a symmetric proximity matrix RANDPROX
% with a zero main diagonal having
% entries that are a random permutation of those in the
% symmetric input proximity
% matrix PROX.
%
% syntax: [randprox] = proxrand_altcomp(prox)
n = size(prox,1);
change = randperm((n*(n-1))/2);
proximity_vector = squareform(prox);
randprox = squareform(proximity_vector(change));
\end{verbatim}
\end{slide}
\normalsize
\begin{slide}{}
To show how these two M-functions work, enter the following:
\bigskip
\begin{verbatim}
prox = targlin(10)
[randprox] = proxrand_altcomp(prox)
\end{verbatim}
\end{slide}
\begin{slide}{Points About M-files}
Functions work in their own workspace --- the Las Vegas property of what goes on in the function space, stays in the function's own workspace. This is different from the use of scripts.
\bigskip
The names of the M-files and the functions should be exactly the same (and meaningful).
\bigskip
Notice the use of the editor, chroma-coding, structured programming,
and so on. This makes writing these M-files much easier.
\end{slide}
\begin{slide}{The Editor}
The MATLAB Editor is a very nice device for writing your functions and scripts, i.e., your M-files.
\smallskip
It is chroma-coded so you can see the special words and commands highlighted in different colors.
\smallskip
There is automatic nested indentation of the statements as you go along to make the programs readable and its structure obvious --- the \verb+end+s line up correctly with the \verb+if+s, \verb+for+s, \verb+while+s, and so on.
\smallskip
There are cute little sounds and colors (or annoying, demanding on your mood) that occur when you begin to commit certain errors.
\smallskip
A debugger operates with the editor, but you don't really need it given the transparency of the written code --- not so in Fortran where you can spend days debugging code.
\end{slide}
\begin{slide}{Graphical User Interfaces (GUIs)}
GUIDE (Graphical User Interface Development Environment) is a layout tool that produces two files: a figure file to hold the actual GUI produced, and a (template) code M-file in which you can program what are called the callbacks, i.e., what happens when a button is pressed or a slider is dragged.
\bigskip
It is generally easiest to modify an existing GUI to do what you want (assuming you can find one).
\bigskip
When you program experiments (and use, perhaps, the Psychophysics Toolbox), you may need to have a standalone MATLAB for accurate timings that doesn't need to go back to any server for license information.
\end{slide}
\begin{slide}{}
We will demonstrate two GUIs --- one demo from the MathWorks is on the traveling salesman problem: \verb+travel.m+;
\bigskip
the second is part of the Statistics Toolbox: \verb+randtool.m+.
\bigskip
We give the ``help'' files; look at the M-files; and run the GUIs.
\end{slide}
\begin{slide}{Programming}
Operational symbols to use (the colons here are not part of the symbol):
\bigskip
\verb+==+ : ``is equal to''
\verb+~=+ : ``is not equal to''
\verb+&+ : ``and''
\verb+|+ : ``or''
\verb+~+ : ``not''
\verb+>=+ : ``greater than or equal''
\verb+<=+ : ``less than or equal''
\verb+>+ : ``greater than''
\verb+<+ : ``less than''
\end{slide}
\begin{slide}{}
Flow control:
\bigskip
if (statement)
\hspace{2ex} \emph{statements}
elseif (statement)
\hspace{2ex} \emph{statements}
else
\hspace{2ex} \emph{statements}
end
\bigskip
if (statement)
\hspace{2ex} (statements)
end
\end{slide}
\begin{slide}{}
for (statement)
\hspace{2ex} \emph{statements}
end
\bigskip
while (statement)
\hspace{2ex} \emph{statements}
end
\end{slide}
\begin{slide}{Data Structures}
Besides a (two-way) row $\times$ column matrix, we have \emph{multidimensional arrays} (e.g., row $\times$ column $\times$ layer) that can be of arbitrary dimensionality.
\bigskip
Cell arrays contain as elements other arrays; these must be accessed with curly braces, ``\verb+{ }+''.
\bigskip
Structures have entries accessed by textual field designators (with a ``.'') --- much like a data-base entity.
\bigskip
Remember that, generally, text used in any context has single quotes: e.g., \verb+ S = 'Larry Hubert' +
\end{slide}
\begin{slide}{}
\tiny
\begin{verbatim}
>> roster(1).name = 'Larry Hubert';
>> roster(1).score = 95;
>> roster(1).grade = 'A';
>> roster(2).name = 'Frieder Koehn';
>> roster(2).score = 100;
>> roster(2).grade = 'A+';
>> roster(3) = struct('name','Michael Regenwetter','score', ...
89,'grade','A-');
>> roster
roster =
1x3 struct array with fields:
name
score
grade
>> roster(3).name
ans =
Michael Regenwetter
\end{verbatim}
\end{slide}
\normalsize
\begin{slide}{MATLAB Speed Tricks}
Vectorization: make sure you use the matrix operations instead of, say, many nested ``for'' loops.
\bigskip
Preallocation: allocate your matrices with, say, \verb+zeros+, of the appropriate size. It is very slow to add to (and expand) the matrices entry by entry.
\bigskip
M-Lint code checker gives a report on your M-files and suggests obsolete code and speed-ups.
\end{slide}
\begin{slide}{}
The Profiler (not the TV show) tells where in a specific M-file the process is spending the most time; if identified, maybe you can reprogram in some way (e.g., vectorize) to increase speed.
\bigskip
There is the program, \verb+matlab2fmex.m+, that produces Fortran code and an interface gateway from a MATLAB M-file and compiles it. When the function is called in MATLAB, the \verb+.dll+ (really, now called \verb+.mexw32+) is used first.
\bigskip
If speed is really important to you, a ten-fold increase is obtained using compiled and callable Fortran routines.
\end{slide}
\begin{slide}{Images}
A two-dimensional array can be displayed as an image where the array elements determine brightness or color of the image (in pixels).
\bigskip
If there are $t$ distinct integers in the array, a \verb+colormap+ is a $t \times 3$ matrix, where the $t$ integers are indexed into the colors.
\bigskip
\verb+imread+ and \verb+imwrite+ can read and write images in the standard formats (\verb+.tiff+, \verb+.jpg+, \verb+.bmp+, and so on).
\end{slide}
\begin{slide}{}
Try this:
\tiny
\begin{verbatim}
>> load durer.mat
>> whos
Name Size Bytes Class Attributes
X 648x509 2638656 double
caption 2x28 112 char
map 128x3 3072 double
>> colormap(map)
>> image(X)
>> axis image
\end{verbatim}
\normalsize
\end{slide}
\begin{slide}{Animation}
We will illustrate one animation of coloring a proximity matrix and seeing how it changes going through row (and column) permutation.
\bigskip
The script is called: \verb+matrixcolor_script.m+, and is given verbatim below. The function \verb+matcolor.m+ is given on the next slide.
\bigskip
\tiny
\begin{verbatim}
[outperm,rawindex,allperms,index] = ...
order(targlin(15),targlin(15),randperm(15),3);
matcolor(targlin(15),allperms,index)
\end{verbatim}
\normalsize
\end{slide}
\begin{slide}{}
\tiny
\begin{verbatim}
function matcolor(datamat,perms,numperms)
%MATCOLOR constructs a color movie of the effects of a series of
% permutations on a proxmity matrix.
% DATAMAT is an $n \times n$ symmetric proximity matrix;
% PERMS is a cell array containing NUMPERMS permutations.
m=moviein(numperms);
for i=1:numperms
pcolor(datamat(perms{i},perms{i}));
axis ij off;
colormap(bone(256));
colorbar;
m(:,i) = getframe;
end
movie(m);
\end{verbatim}
\normalsize
\end{slide}
\end{document}