FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W399991517> ?p ?o ?g. }

• W399991517 abstract "An important research issue in the field of multimedia databases is the retrieval of similar objects. For most applications in multimedia databases, an exact search is not meaningful. Thus, much effort has been devoted to develop efficient and effective similarity search techniques. A widely used approach for implementing similarity search engines is the feature-based approach. In this approach, all multimedia objects stored in a database are transformed into high-dimensional feature vectors, which are then inserted into an index structure to efficiently perform similarity queries. The contribution of this thesis is to explore and propose novel solutions to improve the efficiency of similarity queries in multimedia databases. The thesis begins with a study on how to improve the effectiveness (i.e., the quality of the answer) of a similarity retrieval engine. We first show that by using combinations of feature vectors the effectiveness of the similarity search may be significantly enhanced. Then, we describe methods for computing query-dependent weights to perform linear combinations of feature vectors, which can further improve the effectiveness of the similarity search. As almost all index structures for similarity search developed so far can only deal with single feature vectors, the design and analysis of new index structures is necessary to efficiently perform similarity queries that use combinations of feature vectors. This gives an extra motivation for the techniques studied in the rest of the thesis. In the next part of the thesis, we propose several algorithms and index structures to improve the efficiency of similarity queries. Firstly, we study pivot selection techniques for pivot-based indices. We provide an efficiency criterion based on distance histograms for selecting good set of pivots, and present empirical evidence showing that the technique is effective. Secondly, we describe an improved k nearest neighbor (k-NN) algorithm, which is based on the best-first traversal algorithm proposed by Hjaltason and Samet. Although the original algorithm is already optimal in the number of distance computations, its space requirements are significant. The improved algorithm aims to lower the space requirements by using distance estimators. Thirdly, we present a metric access method for dynamic combinations of feature vectors. The index is pivot-based, and it can take advantage of the previously studied pivot selection techniques. Finally, we introduce an approach that aims to minimize the expected search cost of a similarity query. The idea is to index only the most frequently used combinations of feature vectors. If there are restrictions on the available space for constructing indices, then the resulting optimization problem can be modeled as a binary linear program. As binary linear programs are NP-hard in the general case, we also propose algorithms that quickly find good sets of indices. The last part of the thesis explores the use of graphics processor units (GPUs) for accelerating database operations. We present GPU implementations of a high-dimensional nearest neighbor search and a clustering algorithm. An experimental evaluation shows that the proposed GPU algorithms are an order of magnitude faster than their CPU versions." @default.
• W399991517 created "2016-06-24" @default.
• W399991517 creator A5070507656 @default.
• W399991517 creator A5082317770 @default.
• W399991517 date "2006-01-01" @default.
• W399991517 modified "2023-09-27" @default.
• W399991517 title "Index Structures for Similarity Search in Multimedia Databases" @default.
• W399991517 cites W1487438942 @default.
• W399991517 cites W1492783156 @default.
• W399991517 cites W1501850951 @default.
• W399991517 cites W1529320607 @default.
• W399991517 cites W1530595269 @default.
• W399991517 cites W1532232644 @default.
• W399991517 cites W1541459201 @default.
• W399991517 cites W1554174647 @default.
• W399991517 cites W1561952261 @default.
• W399991517 cites W1564102049 @default.
• W399991517 cites W1566022212 @default.
• W399991517 cites W1566114229 @default.
• W399991517 cites W1588029399 @default.
• W399991517 cites W1592473003 @default.
• W399991517 cites W1601025990 @default.
• W399991517 cites W1603845670 @default.
• W399991517 cites W1633659671 @default.
• W399991517 cites W1660133578 @default.
• W399991517 cites W1660390307 @default.
• W399991517 cites W1672197616 @default.
• W399991517 cites W1723433588 @default.
• W399991517 cites W1754958282 @default.
• W399991517 cites W1770616178 @default.
• W399991517 cites W1782362627 @default.
• W399991517 cites W1946110011 @default.
• W399991517 cites W1971927246 @default.
• W399991517 cites W1977496278 @default.
• W399991517 cites W1983067644 @default.
• W399991517 cites W1991989246 @default.
• W399991517 cites W1994529543 @default.
• W399991517 cites W2004030129 @default.
• W399991517 cites W2005852566 @default.
• W399991517 cites W2011039300 @default.
• W399991517 cites W2016677154 @default.
• W399991517 cites W2021122545 @default.
• W399991517 cites W2024668293 @default.
• W399991517 cites W2026591675 @default.
• W399991517 cites W2028956944 @default.
• W399991517 cites W2028975680 @default.
• W399991517 cites W2032309817 @default.
• W399991517 cites W2038044292 @default.
• W399991517 cites W2046144220 @default.
• W399991517 cites W204885769 @default.
• W399991517 cites W2049644877 @default.
• W399991517 cites W2051793914 @default.
• W399991517 cites W2056162765 @default.
• W399991517 cites W2066799613 @default.
• W399991517 cites W2070572105 @default.
• W399991517 cites W2070705506 @default.
• W399991517 cites W2075597533 @default.
• W399991517 cites W2078321579 @default.
• W399991517 cites W2080115635 @default.
• W399991517 cites W2086345215 @default.
• W399991517 cites W2091967195 @default.
• W399991517 cites W2095897464 @default.
• W399991517 cites W2097921974 @default.
• W399991517 cites W2098912494 @default.
• W399991517 cites W2099789128 @default.
• W399991517 cites W2099824336 @default.
• W399991517 cites W2104594559 @default.
• W399991517 cites W2106626784 @default.
• W399991517 cites W2106642566 @default.
• W399991517 cites W2107278083 @default.
• W399991517 cites W2107373150 @default.
• W399991517 cites W2107627518 @default.
• W399991517 cites W2107779728 @default.
• W399991517 cites W2108157916 @default.
• W399991517 cites W2110449999 @default.
• W399991517 cites W2110764701 @default.
• W399991517 cites W2114640218 @default.
• W399991517 cites W2115135558 @default.
• W399991517 cites W2116466193 @default.
• W399991517 cites W2118269922 @default.
• W399991517 cites W2124222502 @default.
• W399991517 cites W2124561101 @default.
• W399991517 cites W2127218421 @default.
• W399991517 cites W2129430460 @default.
• W399991517 cites W2129905273 @default.
• W399991517 cites W2131074607 @default.
• W399991517 cites W2132540527 @default.
• W399991517 cites W2134624785 @default.
• W399991517 cites W2135306884 @default.
• W399991517 cites W2135346934 @default.
• W399991517 cites W2137565219 @default.
• W399991517 cites W2140190241 @default.
• W399991517 cites W2141579716 @default.
• W399991517 cites W2144473577 @default.
• W399991517 cites W2145087912 @default.
• W399991517 cites W2145725688 @default.
• W399991517 cites W2148008870 @default.
• W399991517 cites W2149708440 @default.
• W399991517 cites W2151135734 @default.
• W399991517 cites W2151855189 @default.

• next