Material Detail

Learning Distance Function by Coding Similarity

Learning Distance Function by Coding Similarity

This video was recorded at 24th Annual International Conference on Machine Learning (ICML), Corvallis 2007. We consider the problem of learning a similarity function from a set of positive equivalence constraints, i.e. "similar" point pairs. We define the similarity in information theoretic terms, as the gain in coding length when shifting from independent encoding of the pair to joint encoding. Under simple Gaussian assumptions, this formulation leads to a non-Mahalanobis similarity function which is effcient and simple to learn. This function can be viewed as a likelihood ratio test, and we show that the optimal similaritypreserving pro jection of the data is a variant of Fisher Linear Discriminant. We also show that under some naturally occurring sampling conditions of equivalence constraints, this function converges to a known Mahalanobis distance (RCA). The suggested similarity function exhibits superior performance over alternative Mahalanobis distances learnt from the same data. Its superiority is demonstrated in the context of image retrieval and graph based clustering, using a large number of data sets.

Quality

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

More about this material

Comments

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