Kernel Methods for Comparing Distributions and Detecting DependenceArthur Gretton, Columbia, 2014 |
|---|
A short course (7 hours) on the theory and application of distribution embeddings, with emphasis on hypothesis testing (two-sample, independence, Lancaster interaction), and on the relation with distance-based test statistics for these problems (Energy Distance, Distance Covariance).
Slides and notes
Talks and slides from the workshop may be found here.