About
On this website you will find some code I wrote during the last couple of years on various projects. Maybe you find something interesting.
Some small packages of mostly matlab source code. They should have a file 'sample_run.m' included where the use is demonstrated with a sample run of the algorithm. Some packages will require the
spider toolbox, a free machine learning toolbox for matlab.
KMEANS Clustering 
Fast code for kmeans clustering based on the paper [Elk03]. This version avoids a lot of distance
calculations by maintaining bounds for all distances. It is a lot
faster than the naive standard implementation of kmeans. Furthermore
at each iteration it produces the same output as the standard
algorithm, so there is no need to use the old one any more.
kmeans@mloss 
EPCA 
Exponential Family PCA (EFPCA or EPCA). EFPCA generalizes Principal Componnent Analysis to the exponential family. The model is taken from [Col02]. Currently it includes only the case of Poisson distributed observed variables. download:epca 
pLSA 
Probabilistic Latent Semantic Analysis. This implementation comes
with both EM and the tempered EM version for Maximum Likelihood
learning. Parts of it are written as mex files so the algorithm
benefits from the sparse structure of the training data. This
implementation follows the paper from [Hof99]
and is based on the code from this ICCV
short course.download:plsa 
LDA 
Multiclass Fisher Linear Discriminant Analysis object for spider. download:lda 
PoEdges 
Product of Edgeperts Image denoising. There is a specific website for this code please visist the PoEdges website for code, examples and images. 
RAP 
The Rate adapting Poisson model for information retrieval and object recognition. This project has its own webpage with code and the datasets. Please visit the RAP website 
References
[Col02] 
A Generalization of Principal Component Analysis to the Exponential Family, M. Collins, S. Dasgupta, R. E. Schapire, Advances in Neural Information Processing Systems 14 
[Hof99] 
Probabilistic Latent Semantic Analysis, T.Hofmann, Proc. of Uncertainty in Artificial Intelligence, UAI'99 
[Elk03] 
Using the Triangle Inequality to Accelerate kMeans, C.Elkan. In Proceedings of the Twentieth International Conference on Machine Learning (ICML'03), pp. 147153.

Contact
pgehler@tuebingen.mpg.de.