||October 13, 2006
||Prof. Alex J. Smola, National ICT Australia
||Maximum Mean Discrepancy for Distribution Testing
|| Hilbert space methods are a promising tool for the analysis
of distribution testing. In particular, they can be used to assess
whether two sets of observations arise from the same distribution.
These tests can be used in the context of data integration for
microarrays, attribute matching, and the correction of sample
bias when test and training sets differ. I will be presenting examples
of their applications and corresponding experimental results.