Material Detail

Bayesian Interpretations of RKHS Embedding Methods

Bayesian Interpretations of RKHS Embedding Methods

This video was recorded at NIPS Workshops, Lake Tahoe 2012. We give a simple interpretation of mean embeddings as expectations under a Gaussian process prior. Methods such as kernel two-sample tests, the Hilbert-Schmidt Independence Criterion, and kernel herding are all based on distances between mean embeddings, also known as the Maximum Mean Discrepancy (MMD). This Bayesian interpretation allows a derivation of optimal herding weights, principled methods of kernel learning, and sheds light on the assumptions necessary for MMD-based methods to work in practice. In the other direction, the MMD interpretation gives tight, closed-form bounds on the error of Bayesian estimators.


  • User Rating
  • Comments
  • Learning Exercises
  • Bookmark Collections
  • Course ePortfolios
  • Accessibility Info

More about this material


Log in to participate in the discussions or sign up if you are not already a MERLOT member.