Abstract
Although there have been many metric access methods (MAMs) developed so far to solve the problem of similarity searching, there is still big need for gapping retrieval efficiency. One of the most acceptable MAMs is M-tree which meets the essential features important for large, persistent and dynamic databases.
M-tree?s retrieval inefficiency is hidden in overlaps of its regions, therefore, its overlaps should be as small as possible. Slim-tree [Traina Jr. et al., 2000] (the M-tree variant), solves this problem by post-processing slim-down algorithm which is too expensive.
We briefly mention our new reinserting algorithm which is dynamic with acceptable construction costs and which reorganizes efficiently index. Another approach how to improve similarity searching has been introduced with the TriGen algorithm [Skopal, 2006] which enables MAMs to perform also non-metric similarity search or faster approximate search.
With the TriGen, far more problems of similarity searching can be solved by MAMs becaus