Here you can find a few links to some implementations or demo apps.
Laplacian Support Vector Machines (LapSVMs) are Semi-Supervised Kernel Machines based on the manifold assumption. The geometrical structure of the marginal distribution is exploited to enhance the quality of the classifier.
I have developed a Matlab library to solve the primal formulation of the LapSVM problem. Thanks to some stability-based early stopping conditions, training times and complexity of the training algorithm have been significantly reduced with respect to the original dual formulation. See
- S. Melacci and M. Belkin. Laplacian Support Vector Machines Trained in the Primal. Journal of Machine Learning Research, 12(3):1149–1184, 2011
For all details (and to download the library), visit my webpage dedicated to Training LapSVMs in the Primal.
Active Appearance Models
Active Appearance Models (AAMs) can compaclty describe the shape and the texture of a variety of objects, allowing us to generate new instances or to match (and track) objects of the same "category". In the case of human faces, here you can find a simple Java Applet based on my implementation of AAMs from
- T. F. Cootes, G. J. Edwards, and C. J. Taylor. Active appearance models. IEEE TPAMI, 23(6):681–685, 2001
The sample model has been estimated from 30 faces and it is explained by just 20 modes (only the first 12 ones are shown). In order to reduce the physical dimension of the model file, the generated texture is really small, so many details are lost when zooming in.
Click on the following button (it will be visible only if you have a Java Runtime Engine (JRE 1.6 or an earlier version) installed in your system) to run the demo, then use the File menu to download the sample model (warning: the download may take a while).