	Department of Engineering

University of Cambridge > Engineering Department > computing help > programs

Matlab

Contents

Introductions - Help
Local articles - Articles elsewhere
Toolboxes - Simulink - COMSOL
Using C++ and F77 - Optimisation - Parallelisation
Installing - Price List - Student Version
Printable documentation
Local updates (last changes September 2009)

MATLAB is a high performance interactive software package for scientific and engineering computation. MATLAB integrates numerical analysis, matrix computation, signal processing and graphics in an easy-to-use environment where problems and solutions are expressed just as they are written mathematically. To start MATLAB on our system just type matlab in a terminal window or find the Matlab icon in the applications panel (use the greenarrow button).

Introductions

We have various handouts online. Mathworks' own Getting Started with MATLAB is useful. Alternatives are CUED's Getting Started with Matlab (PDF version) and Matlab by Example. The Matlab Primer gives an out of date description of facilities. Using Matlab at CUED (PDF version) lists the added local features, and gives answers to many questions commonly asked at CUED. Mathworks' Using MATLAB gives more details on maths, graphics, object-oriented programming etc.

A Matlab databook (PDF version) is in preparation, aimed at 3rd and 4th years.

Toolboxes

Toolboxes are libraries of extra routines.

The Symbolic Math Toolbox gives access to the full Maple library of routines to factorise, integrate, solve etc, symbolically. Type mhelp intro within matlab for details.
The Statistics Toolbox User's Guide includes a tutorial.
Also installed is the WaveLab toolbox for investigating Wavelets. Read the introductory handout for details. Type WavePath to get started and see a list of demos.

Typing ver inside Matlab lists the other toolboxes that we're licensed to use. If we don't have something you want, try

Getting help

Matlab has its own text-only help system (type help from the matlab prompt) and WWW-based documentation (see the Help Desk or type doc). There are some video tutorials you can play (see the Demos sections, especially "Desktop Tools and Development Environment"). Use matlabdoc -a to list documentation files available in the alternative (PDF) format. "matlabdoc -a newfeat", for example, will display the 'New Features' document.

More answers can be obtained from the Mathworks' search facility including

Using Movie and animation guide

Mathworks (their european mirror site may be faster) also maintains a Support and Customer Service page which includes a facility that lets you search for answers by typing in a keyword, and a MATLAB In Education page that leads to various examples and courses. Some short local articles are now online

Input/Output - File Input and Formatted Output - Spreadsheets with Matlab - Experimental data with Matlab
Language Features - Object-orientated programming - Finding Matlab functions - Function handles - Matlab: higher dimensions - Matlab: NaN and Inf - Matlab cells and structs
Special Topics - Curve fitting - Symbolic Maths Toolbox - Matlab and ODEs - Matlab and 1B dynamics for CUED students - Mex files - Find local min and max values
Optimising - Vectorisation Tricks - Faster Scripts
Customising - Matlab's desktop

Articles elsewhere include

Techniques for Debugging MATLAB M-files (Mathworks)
Fixed-Point Modeling in an Ultra Wideband (UWB) Wireless Communication System (Mathworks)
Calculating Complex Integrals (Mathworks)
Using MATLAB's meshgrid command and array operators (Mathworks)
Exporting Figures for Publication (Mathworks)
Making Pretty Graphs (Mathworks)
Troubleshooting Common Floating-Point Arithmetic Problems (Mathworks)
Support Guide to Graphics Rendering and Troubleshooting
The Technical Support Guide to Printing (Mathworks)
The Technical Support Guide to Memory Management (Mathworks)
Avoiding Out of Memory Errors (Mathworks)
GUI Building with Handle Graphics and GUIDE (Mathworks)
File I/O Guide (Mathworks)
Curve Fitting Guide (Mathworks)
Matlab Snippets
Introduction to Object-Oriented Programming in MATLAB

Numerical Computing with MATLAB and Experiments with MATLAB by Cleve Moler are online textbooks.

If you have problems, you can try using matlab's debug facility (begin with dbstop). Alternatively, local users can look at CUED's Matlab FAQ, mail Tim Love, look at the global newsgroup comp.soft-sys.matlab (a WWW interface is at newsreader.mathworks.com). A list of the newsgroup's Frequently Asked Questions (FAQ) is online.

Some ancient manuals can be borrowed from the operators in the Engineering dept's DPO. Documentation from other universities is available too.

Some Control Tutorials from the University of Michigan may be useful. They include animations.
The University of Westminster has some online matlab courses
DSP-related Matlab facilities
The University of New Hampshire has online tutorials
Southern Illinois University have MATLAB Tutorials on Linear Algebra, Numerical Analysis, etc.

Simulink

SIMULINK is an extension of MATLAB that lets you model a system graphically. It adds to Matlab many features specific to dynamic systems while retaining all of MATLAB's general purpose functionality. See Simulink's user guide for details.

In January 2007 Mathworks identified bugs in R14SP3 Simulink(R) code generation products, which in rare instances generate incorrect code that is not easily detected. These bugs have been fixed in subsequent releases: R2006a, R2006b, or the upcoming R2007a release, so if you used R14SP3 to generate code, upgrade and regenerate.

COMSOL

COMSOL is a FE package. Type "comsol" to run it, or click on its icon. See our COMSOL page for further information. email amb2 with any questions.

Using Matlab with Fortran and C++

Matlab can also be used from C++ and fortran as a library of maths and graphics routines. Also C++ and fortran code can be called from within matlab. An Interface Guide is online. Current examples are in /usr/local/apps/matlab/matlabR2007a/extern/examples/ on our Linux servers. Local users using these files on the linux servers should note that some local configuring may be required - see Matlab: configuring mex page.

We have 1 license centrally maintained for Mathworks' compiler that converts MATLAB programs into stand-alone applications (though the royalty-free MATLAB Compiler Runtime (MCR) needs to be installed for the "stand-alone application" to work). Note that MATLAB's matrix-oriented operations are already highly optimized. As a result, the MATLAB Compiler is unlikely to significantly speed up functions that spend most of their time in MATLAB's built-in indexing, math, or graphics functions, or that are heavily vectorized.

Optimisation and long jobs

See the GridEngine documentation before running long jobs. Matlab scripts can often be speeded up considerably with little work - see the Faster Scripts document. The first 2 methods mentioned below have speeded up some CUED programs by a factor of 10 or so.

Use sparse matrices when you can. They may save you a great deal of space and time. See matlab's sparse command.
Use matrix operations instead of for loops - i.e. vectorise your code. See the Tricks page and Peter Acklam's MATLAB array manipulation tips and tricks.
Functions are faster than scripts - they get compiled internally the first time they are run.
If you know the maximum size that a matrix will be, create it that size initially, rather than letting it grow incrementally.
Use pack every so often to tidy up memory usage.

Parallelisation

The Matlab and parallel computing page has information. Newer versions of matlab exploit multi-processors internally. If you run MATLAB R2007a or newer on a multiple-CPU system (multiprocessor or multicore), you can enable "multithreaded computation". It's not set by default but you can set it in Preferences

On our linux servers (which have 4 CPUs, so the 'Automatic' option sets the number of threads to 4) the following graph shows the speed-up factors for a range of tasks multithread

Note that some operations aren't speeded up, whereas others are over 3 times faster (the y axis running from 0 to 3.5). To exploit multi-processors more fully (or to use several machines in parallel) you need extra software. Here are some of the alternatives, not all of which are free.

MatlabMPI (used at CUED. Offers low-level multi-core/processor facilities, and multi-computer facilities if the machines share the same filesystem)
pMatlab: Parallel Matlab Toolbox (uses MatlabMPI)
multicore (partly superceded by Matlab 2007a)
Parallel Computing Toolbox (Mathworks - to use several cores/processors in a more sophisticated way than the default allows)
Distributed Computing Server (Mathworks - to use several machines)
MATLAB plug-in for CUDA (CUDA is a library that uses an nVidia board as a number-cruncher)

See Mathworks' Enhancing Multicore System Performance Using Parallel Computing with MATLAB article.

New Facilities

Added facilities include

September 09: The default matlab is release R2009b.
November 08: movietogif installed - converts movies to animated GIFs.
September 08: The default matlab is release R2008a.
April 08: matlabR2008a installed. Default is still the old matlab.
July 07: Matlab R2007a installed on the Unix Teaching System.

Installing Matlab

Our site-license covers covers the following types of machine within the department - HP, Apple OSX, Compaq (DEC), SUN, Linux RedHat and machines running 32-bit/64-bit versions of Microsoft Windows which are connected to the department's network.

See the On-line Installation of Windows version of MATLAB page for Windows machines.

Mail other installation requests to pc-support or research-support. Staff who need stand-alone versions of MATLAB should contact The Mathworks direct.

It's possible to have CUED's Matlab installed on your own machine as long as the machine is connected to the internet when matlab is used. See

See our MATLAB Licensing Information and MATLAB - Frequently Asked Questions on Licensing at CUED for licensing details.

Old versions of Matlab are installed on CUED's central system under /usr/local/apps/matlab/

Purchasing Additional Licences for ToolBoxes

Mail amb2 for license details.

Price List for Matlab

MATLAB Academic Price List.

Student Version (Buy Your Own Copy)

Students who need MATLAB on desktops/laptops (non-networked) can order it from Bookshops (this includes a book, Simulink and some Symbolic Toolbox functions). See Mathworks' Student Version page for details.

Any additional toolboxes (priced at US$59) can be purchased directly from the American web site using a credit card. These are only for bona fide students.

octave is free and may be useful to you. It's used in CUED undergraduate courses and is on the MDP Resource disc.

| computing help |