Title :
On the optimality of the hash-coding type nearest neighbour search algorithm
Author :
Aslanyan, Levon H. ; Danoyan, H.E.
Author_Institution :
Inst. for Inf. & Autom. Problems of NAS RA Yerevan, Yerevan, Armenia
Abstract :
The algorithmic problem of nearest neighbour search and the earlier proposed hash-coding algorithm (Elias algorithm) are considered. The problem of optimality characterization depending on the geometrical structure of blocks of the algorithm in a balanced hash-coding scheme is considered. As a supportive technique for this the structure of consecutive neighbourhood layers to the subsets of binary cube is considered. Using the known properties of isoperimetric sets it is proved that the algorithm is optimal when mentioned blocks are isoperimetric sets.
Keywords :
file organisation; geometry; search problems; set theory; Elias algorithm; algorithmic problem; balanced hash-coding scheme; binary cube subsets; consecutive neighbourhood layers; geometrical structure; hash-coding type nearest neighbour search algorithm optimality; isoperimetric set properties; optimality characterization problem; Algorithm design and analysis; Complexity theory; Finite element analysis; Heuristic algorithms; Machine learning algorithms; Standards; Vectors; Best-match; Elias algorithm; hash-coding schema; isoperimetricset; nearest neighbour search;
Conference_Titel :
Computer Science and Information Technologies (CSIT), 2013
Conference_Location :
Yerevan
Print_ISBN :
978-1-4799-2460-8
DOI :
10.1109/CSITechnol.2013.6710336