I want to perform K-Nearest Neighbor Search in multidimensional space but not using for example L2 distance but I want user to specify some "similar" pairs-examples and then perform search using this information.

what algorithm I can use for this?

asked Nov 20 '12 at 07:51

mrgloom's gravatar image

mrgloom
91131519

You could have a look at something like Neighbourhood Components Analysis (NCA) or Large Margin Nearest Neighbour (LMNN) to map your data into a low-dimensional space where those example pairs are close together. You can then use L2 distance in the new space.

(Nov 20 '12 at 08:58) Sander Dieleman
Be the first one to answer this question!
toggle preview

powered by OSQA

User submitted content is under Creative Commons: Attribution - Share Alike; Other things copyright (C) 2010, MetaOptimize LLC.