So far most the work for my PhD has been centerd around the unsupervised learning of invariances based on Slow Feature Analysis (SFA). More specifically we build a model for invariant object recognition. This work was done in the group of Laurenz Wiskott together with Mathias Franzius (who previously used SFA for a place cell model). Some examples of the visual stimuli we used are shown below:

fish image fish image fish image fish image

Our model consists of a hierarchical network of SFA units. I have integrated all the required parts into the Modular toolkit for Data Processing (MDP), an open source library for the Python programming language.