Programming Resources

PRTools

PRTools is a toolbox for MATLAB, developed at DELFT in the Netherlands. Both the Kuncheva text and the van der Heijden et al. textbook use a number of PRTools-based examples. In particular, the van der Heijden text was written by the group that developed PRTools, and this text is available for free to the RIT community (see the list of books on the Other Resources page).

PRTools comes with a nice overview and reference manual given as a .pdf in the prtools_ac directory. The overview, summary of built-in examples, and reference tables in Section 7 are particularly useful. When installing PRTools, make sure to add it to your MATLAB path (from the File menu, click: File -- Set Path). For those new to MATLAB, there are a number of tutorials online, and MATLAB comes with a large set of demonstrations (look under the 'Help' menu). You do not need to purchase MATLAB, as the computer science department has MATLAB running on the lab machines.

Tutorial/Examples: There is a set of exercises giving example uses of PRTools available online here. A number of datasets needed for the examples (and for your own experimentation) are available here.

Examples

• Clustering Banana-shaped data sets using k-means clustering: clustereg.m
• Generating samples of Gaussian (normal) distributions and plotting them in MATLAB: gdisteg.m
• Creating, displaying, and testing a linear discriminant classifier (LDC: classes assumed to be normally distributed with identical covariance matrices): lineareg.m
• Using the Quadratic Discriminant Classifier (QDC: Classes assumed to be normally distributed, but with different covariance matrices): quadeg.m
• A similar example, but using k-Nearest Neighbor classification: nn.m
• Voronoi diagrams (illustrating decision regions for 1-NN, points in the plane closest to each training sample):
  • voronoieg.m
  • voronoieg2.m

OpenCV

The Open Source Computer Vision Library, is a C/C++ library for constructing fast, real-time computer vision programs. It includes a large number of image processing, pattern recognition, and machine learning algorithm implementations. Documentation for the most recent OpenCV release (2.1) is available for the following languages:

• C
• C++, and
• Python

Using Python scripts is probably the fastest way to start playing with the library, as you can use it interactively within a Python shell.

A brief overview of OpenCV has been provided by Gady Agam (Introduction to programming with OpenCV). A book on OpenCV, Learning OpenCV is available, with a number of interesting examples. It can be obtained (free!) as an electronic book through the online RIT library catalogue.

Please note that the library underwent a significant revision recently, so you may find conflicting syntax and different approaches when you compare source code examples.