Rules
The final part of the programming tutorial consists of a project, which is mandatory (if you want the certificate for this lecture). The projects are more research oriented than the problem sets. The projects are assigned to teams of two (or three if there is an odd number of participants). At the end of the project you have to give a presentation talk and your source code will be evaluated as well.
To make sure that the projects are successful it is expected that you meet me (Niko) once a week (or if that is not possible at least tell me about your progress via email). I will also report the progress to Prof. Laurenz Wiskott. The default date for these meetings would be during the time of the programming tutorial. Alternatively we can try to find another date, but we should not end up with a different date for each team.
Update: The presentation talks for the projects will be on the 23rd of Jule. On the 2nd and the 9th of Jule, Prof. Laurenz Wiskott will attend the project meetings as well.
The talks each should take 13 minutes, with 7 minutes for discussion afterwards. You should try hard not to exceed the 13 minutes, since not finishing a talk on time is generally considered as bad style. On the other hand you should really use your 13 minutes and not finish too early.
Project Topics
There are four project topics, each team can pick one (first come, first serve). Projects which have been assigned to a team are marked as such. To get a project assignment you can contact me (via email or come to my office).
 SelfOrganizing Maps (assigned) in the visual system (SOM). This topic is based on the corresponding part of the lecture with vector quantization. One goal would be to create realistic maps for position, orientation and ocular dominance.
 Bayesian Network (assigned) with continuous variables. This project is based on this paper on attention in the visual system. You would implement a Bayesian network, with continuous variables. We might actually use your work for a similar research project in our group, therefore the implementation should be in Python. The reference for the paper mentioned above is:

Comparison of two weighted integration models for the cueing task: linear and likelihood
SS Shimozaki, MP Eckstein, CK Abbey  Journal of Vision, 2003  journalofvision.org
 Hopfield Networks (assigned). In this project you will investigate the properties of Hopfield networks, e.g. storage capacity and attractors. Eventually you could also investigate applications in the hippocampus.
 Complex Cells from SFA (assigned). In this project you would investigate how SFA can be used to understand complex cell properties in the visual cortex. The foundation for this project is this research project. Some relevant publications are (which can be downloaded from Laurenz Wiskott's homepage):
 Berkes, P. and Wiskott, L. (1. February 2007).
Analysis and interpretation of quadratic models of receptive fields
Nature Protocols 2(2):400407.
(bibtex, abstract.html)
 Berkes, P. and Wiskott, L. (20. July 2005).
Slow feature analysis yields a rich repertoire of complex cell properties
Journal of Vision, 5(6):579602, http://journalofvision.org/5/6/9/, doi:10.1167/5.6.9.
(bibtex, abstract.html, paper, paper.pdf)
 Wiskott, L. and Sejnowski, T.J. (April 2002).
Slow feature analysis: Unsupervised learning of invariances
Neural Computation, 14(4):715770.