This workshop asks for insights how far we may/can push the theoretical boundary of using data in the design of learning machines. Can we express our classification rule in terms of the sample, or do we have to stick to a core assumption of classical statistical learning theory, namely that the hypothesis space is to be defined independent from the sample? This workshop is particularly interested in – but not restricted to – the ‘luckiness framework’ and the recently introduced notion of ‘compatibility functions’ in a semi-supervised learning context (more information can be found at http://www.kuleuven.be/wehys).

This seems interesting. I (with my advisor) have proposed a framework in studying classification algorithms where the space of classifiers is induced by the learning algorithm and the underlying joint distribution. The details can be seen in the papers (recently accepted in JMLR and TKDD) uploaded on http://www.cise.ufl.edu/~asd/Research.html. The framework essentially considers studying the moments of the GE error over this induced space of classifiers. Since, this space can be potentially much smaller than the space considered in SLT for e.g., the results are tighter. A lot of consideration is also given to scalability of the approach too. Though much work needs to be done in extending the analysis and making it more scalable, I think it is a novel way of looking at the problem and can be of interest considering the workshop topic. What say you?

[...] to Kristiaan Pelckmans for posting about the upcoming Workshop on Empirical Hypothesis Spaces at NIPS 2008: This workshop asks for [...]

Hereby we’d like to invite you at our workshop on “New Challenges in Theoretical Machine Learning: Learning with Data-dependent Concept Spaces”, formerly called “Learning over Empirical Hypothesis Spaces”. Note that the exact date is now *Friday 12 December* in the *Hilton: Sutcliffe A*. You can find the tentative schedule at http://www.kuleuven.be/wehys/. We have still room to accommodate an extra spotlight and poster presentation. In case of questions do not hesitate to contact Wehys08@gmail.com.

Nina, Shai, Avrim, John and Kristiaan