9 Jan 2004
Projected Nearest Neighbor Problem
Speaker: David CHENG
Abstract
K-nearest neighbor (k-NN) problem has many applications, such as in
multimedia search, fraud detection and information retrieval. However,
most current solutions on this problem perform poorly when the dimension
of the dataset is high. In my presentation, I would first discuss the
issues of the k-NN problem in high dimensional space. Then I would
introduce a very similar problem called projected k-NN problem. I would
also study some previous method on this problem as well as presenting our
proposed method. Our method would be mainly based on the data structure
called B+-Tree.
|