Matthew Roughan

Various bits of code that may be useful to people:

• Matlab (most of this is also hosted on Matlab Central)
  • crosshatch_poly.m lets you do lines to fill a convex polygon (rather than filling the polygon with a flat color using the Matlab fill.m function. For instance see the following picture
    
    which was generated using crosshatch_poly_test.m.
  • plot_feasible.m is a simple bit of code for visualizing 2D linear programming problems. Its primarily of use for teaching Linear Algebra. It generates pictures of the form of
    which was generated using plot_feasible_test.m, though there are many options that can be used to alter the output. Note it uses the optimization toolbox, though this can probably be removed in the future.
  • ipf.m does Iterative Proportional Fitting (IPF). There are a couple of ways to use IPF, but in this case the aim is to find a matrix close to the input T that satisfies contraints on its row sums and column sums (its marginals). In particular we aim to find S with row sums equal to R, and column sums equal to C. To see how to use it have a look at ipf_test.m. More information can be found at Wikipedia or various other places on the web.
    There is a second version of this code used in our traffic matrix work, which is designed to allow one to find a matrix S that satisfies an arbitrary set of linear constraints y=Ax on a vectorized version x of the matrix T. Its not quite in a nice form to hand out, but it you need it ask me.
  • gaussian_mixture_model.m does simple 1D Gaussian Mixture Model estimation using the EM algorithm. It uses a couple of little functions for estimating Gaussian distributions, so best to grab the zip file. The result is that we can break a multimodal distribution such as displayed below into a random mixture of Gaussians.
    
  • randp.m generates Pareto random variables (of type I). The Pareto distribution is a classic heavy-tailed or power-law distribution. See "Statistical Distributions", Evans, Hastings and Peacock, Wiley, 1993 or my distributions page or Wikipedia for more information. For an example see the following generated using randp_test.m
    
  • Some useful bits of networking code: to come
  • Traffic matrix generation: to come
  • IGP weight generation: to come
• Perl
  • Traffic decomposition code. to come

Matthew Roughan

Last modified: Mon Aug 3 11:30:21 2009