A MATLAB tutorial: HTML, PostScript

Decision tree for optimization software

MATLAB code

• Ellipse plot
• Lightspeed Matlab toolbox for fast and stable implementation of many common routines.
• NetLab software implements a number of techniques discussed in class.
• Tips for tuning your MATLAB code. The vectorization tips at the bottom will give the most speed-ups.

C++ code

• My image libraries. If you want to load and save with PNG files (recommended), you'll need to link with libpng and #define USE_LIBPNG.
• My matrix libraries that provide a wrapper for BLAS and LAPACK. Here's an example code fragment. On UNIX, you'll need libblas and liblapack (which should generally be installed already; if not, you can probably find an RPM or equivalent). On Windows, you can use the LIB/DLLs files provided here.
• 2D symmetric eigenvector decomposition (useful for inverting covariance matrices). (You don't need this if you're using the above matrix libraries).
• Hybrid Monte Carlo code. The matrix libraries are also required.

External programming resources

• L-BFGS, a constrained numerical solver that is very efficient for when you can compute gradients. It's written in Fortran.
  • The library translated to C using f2c:bfgs.c, f2c.h.
  • An high-level C++ wrapper class: BFGSwrap.h BFGSwrap.cpp, example.cpp
  • Marge's wrapper in C: C example, Makefile.
• MINPACK, another excellent Fortran numerical solver. I haven't used it personally, so I don't have any wrapper code for it, although it's relatively simple to write it if you know how.