A Hashed Schema for Similarity Search in Metric Spaces
| Authors | |
|---|---|
| Year of publication | 2000 |
| Type | Article in Proceedings |
| Conference | Proceedings of the First DELOS Network of Excellence Workshop on "Information Seeking, Searching and Querying in Digital Libraries" |
| MU Faculty or unit | |
| Citation | |
| Field | Information theory |
| Description | A hashing schema for similarity search in generic metric spaces is investigated, assuming that only distances for pairs of objects are known. Similarity Hashing partitions data objects in bounding regions without overlapping. The proposed structure aims at reducing both the I/O and the CPU search costs. Contrary to the traditional tree-based approaches, specific upper-bounds on the search cost can be determined and the data organized in such way that the I/O costs never exceed those needed for sequential scan. Though the current version is static, it can be modified for dynamic data; it is also suitable for parallel implementations. Insertion is fast, and once the computed distances in the search phase are reused to significantly reduce the number of distance computations, that is proportional to the CPU costs. Experiments with the current prototype provide very encouraging results, especially for small similarity ranges. |
| Related projects: |