Approximate k-NN Search in High-Dimensional Spaces Based on the Clustering Index Tree
Yingjie Kuang · Modern Computer · 2007
In this paper,we present an approximate k-NN search method in high-dimensional spaces which is based on clustering index tree.We give a detail description of the build tree algorithm and the k-NN search algorithm.Due to the limit of improvement in k-NN search efficiency by the traditional index structure,we achieve much higher efficiency at the cost of little precision by combining the approximate search and clustering index tree.Experimental results indicate that the precision loss is very small and the search speed outperforms other index methods.