Abstract
Distance permutation indexes support fast proximity searching in high-dimensional metric spaces. Given some fixed reference sites, for each point in a database the index stores a permutation naming the closest site, the second-closest, and so on. We examine how many distinct permutations can occur as a function of the number of sites and the size of the space. We give theoretical results for tree metrics and vector spaces with L1, L2, and L[infinity] metrics, improving on the previous best known storage space in the vector case. We also give experimental results and commentary on the number of distance permutations that actually occur in a variety of vector, string, and document databases.
| Originalsprog | Engelsk |
|---|---|
| Tidsskrift | Journal of Discrete Algorithms (Amsterdam) |
| Vol/bind | 7 |
| Udgave nummer | 1 |
| Sider (fra-til) | 49-61 |
| Antal sider | 13 |
| ISSN | 1570-8667 |
| DOI | |
| Status | Udgivet - 2009 |
| Udgivet eksternt | Ja |
Emneord
- Distance permutation
- Metric Space
- Nearest neighbour
- Voronoi diagram