Material Detail

Kernel Learning for Novelty Detection

Kernel Learning for Novelty Detection

This video was recorded at NIPS Workshop on Kernel Learning: Automatic Selection of Optimal Kernels, Whistler 2008. We consider kernel learning for one-class Support Vector Machines. We consider a mix of 2- and 1-norms of the individual weight vector norms allowing control of the sparsity of the resulting kernel combination. The resulting optimisation can be solved efficiently using a coordinate gradient method. We consider an application to automatically detecting the appropriate metric for a guided image search task.


  • 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.