Trading Quality for Time with Nearest-Neighbor Search

Weber, Roger; Böhm, Klemens

doi:10.1007/3-540-46439-5_2

Cited by 59 publications

(50 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The use of efficient hierarchical multi-dimensional access methods with optimization techniques in the processing of index nodes is a very interesting choice (since the filtering power can be improved notably). Future research may include the use of approximation techniques on VA-files [19,8], the cost estimation of VA-filebased DJQ [19] and the study of the buffering impact over these DJQs, as in [4].…”

Section: Discussionmentioning

confidence: 99%

VA-Files vs. R*-Trees in Distance Join Queries

Corral

D'Ermiliis

Manolopoulos

et al. 2005

Advances in Databases and Information Systems

View full text Add to dashboard Cite

Abstract. In modern database applications the similarity of complex objects is examined by performing distance-based queries (e.g. nearest neighbour search) on data of high dimensionality. Most multidimensional indexing methods have failed to efficiently support these queries in arbitrary high-dimensional datasets (due to the dimensionality curse). Similarity join queries and K closest pairs queries are the most representative distance join queries, where two highdimensional datasets are combined. These queries are very expensive in terms of response time and I/O activity in case of high-dimensional spaces. On the other hand, the filtering-based approach, as applied by the VA-file, has turned out to be a very promising alternative for nearest neighbour search. In general, the filtering-based approach represents vectors as compact approximations, whereas by first scanning these approximations, only a small fraction of the real vectors is visited. Here, we elaborate on VA-files and develop VA-file based algorithms for answering similarity join and K closest pairs queries on highdimensional data. Also, performance-wise we compare the use of VA-files and R*-trees (a structure that has been proven to be of robust nature) for answering these queries. The results of the comparison do not lead to a clear winner.

show abstract

Section: Discussionmentioning

confidence: 99%

VA-Files vs. R*-Trees in Distance Join Queries

Corral

D'Ermiliis

Manolopoulos

et al. 2005

Advances in Databases and Information Systems

View full text Add to dashboard Cite

show abstract

“…PCA is an unsupervised method that maximally preserves the variance of the data, and LDA is a supervised method that achieves maximal class separation by maximizing the ratio of between-class variance to the within-class variance. In approximate search techniques based on Vector Approximation (VA)-Files [18], dimensionality reduction is obtained by quantizing the original data objects. Other techniques that fall in the category of space transformation are FastMap [19], mainly used in vector spaces, and MetricMap [20] suitable for metric spaces.…”

Section: Background and Related Workmentioning

confidence: 99%

Multimedia search and retrieval using multimodal annotation propagation and indexing techniques

Lazaridis

Αξενόπουλος

Rafailidis

et al. 2013

Signal Processing: Image Communication

View full text Add to dashboard Cite

In this paper, a novel framework for multimodal search and retrieval of rich media objects is presented. The searchable items are media representations consisting of multiple modalities, such as 2D images, 3D objects and audio files, which share a common semantic concept. A manifold learning technique based on Laplacian Eigenmaps was appropriately modified in order to merge the low-level descriptors of each separate modality and create a new low-dimensional multimodal feature space, where all media objects can be mapped irrespective of their constituting modalities. To accelerate search and retrieval and make the framework suitable even for web-scale applications, a multimedia indexing scheme is adopted, which represents each object of the dataset by the ordering of a number of reference objects. Moreover, the hubness property is introduced in this paper as a criterion to select the most representative reference objects, thus, achieving the maximum possible performance of indexing. The content-based similarity of the multimodal descriptors is also used to automatically annotate the objects of the dataset using a predefined set of attributes. Annotation propagation is utilized to approximate the multimodal descriptors for multimodal queries that do not belong to the dataset.

show abstract

“…In [9,11,19] dimensionality reduction techniques are used where lowerbounding rules are ignored when dismissing dimensions and the focus is only on preserving close approximations of distances. In [34] the authors used VA-files [35] to find nearest neighbors by omitting the refinement step of the original exact search algorithm and estimating approximate distances using only the lower and upper bounds computed by the filtering step. Finally, in [30] the authors partition the data space into clusters and then the representatives of each cluster are compressed using quantization techniques.…”

Section: Related Workmentioning

confidence: 99%

Query-sensitive embeddings

Athitsos

Hadjieleftheriou

Kollios

et al. 2007

ACM Trans. Database Syst.

View full text Add to dashboard Cite

A common problem in many types of databases is retrieving the most similar matches to a query object. Finding those matches in a large database can be too slow to be practical, especially in domains where objects are compared using computationally expensive similarity (or distance) measures. This paper proposes a novel method for approximate nearest neighbor retrieval in such spaces. Our method is embedding-based, meaning that it constructs a function that maps objects into a real vector space. The mapping preserves a large amount of the proximity structure of the original space, and it can be used to rapidly obtain a short list of likely matches to the query. The main novelty of our method is that it constructs, together with the embedding, a query-sensitive distance measure that should be used when measuring distances in the vector space. The term "querysensitive" means that the distance measure changes depending on the current query object. We report experiments with an image database of handwritten digits, and a time-series database. In both cases, the proposed method outperforms existing state-of-the-art embedding methods, meaning that it provides significantly better trade-offs between efficiency and retrieval accuracy.

show abstract

Trading Quality for Time with Nearest-Neighbor Search

Cited by 59 publications

References 10 publications

VA-Files vs. R*-Trees in Distance Join Queries

VA-Files vs. R*-Trees in Distance Join Queries

Multimedia search and retrieval using multimodal annotation propagation and indexing techniques

Query-sensitive embeddings

Contact Info

Product

Resources

About