My main research is directed towards Learning Classifier Systems (especially
XCS and its derivatives), a family of versatile evolutionary rule-based
machine learning algorithms. One of the main arguments for their application
is that they generate models that are more easily interpreted than the
models generated by other learning systems such as artificial neural
networks. However, we still don't yet fully understand in a formal way
how LCS work, that is, which kinds of problems LCS are able to learn exactly
and why they can learn them—which weakens the argument for them by
quite a bit. Through my work, I want to develop a more formal understanding
of LCS and finally prove several assumptions that were, up to now, only
validated experimentally.
Other than that, I'm very interested in
-
reinforcement learning
-
evolutionary machine learning algorithms
-
functional programming, especially using Haskell
-
ways to incoporate a more formal (functional) view into everyday
software such as operating systems (NixOS) or window managers (XMonad)
- free software
