Reduction of Distance Computations in Selection of Pivot Elements for Balanced GHT Structure
László Kovács · Transactions on machine learning and data mining · 2013
In general metric spaces, one of the most widely used indexing techniques is the partitioning of the objects using pivot elements. The efficiency of partitioning depends on the selection of the appropriate set of pivot elements. In the paper, some methods are presented to improve the quality of the partitioning in GHT structure from the viewpoint of balancing factor. The main goal of the investigation is to determine the conditions when costs of distance computations can be reduced. We show with different tests that the proposed methods work better than the usual random and incremental pivot search methods.