ΠΟΛΥΤΕΧΝΕΙΟ ΚΡΗΤΗΣ
Τμήμα Ηλεκτρονικών Μηχανικών & Μηχανικών Υπολογιστών
ΠΑΡΟΥΣΙΑΣΗ ΜΕΤΑΠΤΥΧΙΑΚΗΣ ΕΡΓΑΣΙΑΣ
ΑΡΓΥΡΙΟΥ ΜΙΧΑΗΛ
με θέμα
“Αναζήτηση κοντινότερου γείτονα σε μη ισοζυγισμένα κατανεμημένα tries με brand-and-bound αναζήτηση”
“Branch-and-bound nearest neighbor searching over unbalanced trie-structured overlays”
Searching is one of the fundamental problems in Computer Science. The hype is searching huge data volumes and therefore p2p frameworks have been built the last years. An interesting type of search is nearest neighbor search where a query is given and the k most similar data are returned. The distance function and the type of data vary for each domain space. In a previous work we had crafted GRaSP a novel DHT framework for generalized range search over p2p overlays based on unbalanced tries. In this paper we extend it to support k-NN queries. We prove the searching algorithm’s good behavior theoretically and experimentally.
