Showing papers by "Nikos Mamoulis published in 2001"

Multiway spatial joins

Abstract: Due to the evolution of Geographical Information Systems, large collections of spatial data having various thematic contents are currently available. As a result, the interest of users is not limited to simple spatial selections and joins, but complex query types that implicate numerous spatial inputs become more common. Although several algorithms have been proposed for computing the result of pairwise spatial joins, limited work exists on processing and optimization of multiway spatial joins. In this article, we review pairwise spatial join algorithms and show how they can be combined for multiple inputs. In addition, we explore the application of synchronous traversal (ST), a methodology that processes synchronously all inputs without producing intermediate results. Then, we integrate the two approaches in an engine that includes ST and pairwise algorithms, using dynamic programming to determine the optimal execution plan. The results show that, in most cases, multiway spatial joins are best processed by combining ST with pairwise methods. Finally, we study the optimization of very large queries by employing randomized search algorithms.

Selectivity Estimation of Complex Spatial Queries

Abstract: Several studies have focused on the efficient processing of simple spatial query types such as selections and spatial joins. Little work, however, has been done towards the optimization of queries that process several spatial inputs and combine them through join and selection conditions. This paper identifies the dependencies between spatial operators and illustrates how they can affect the outcome of complex queries. A thorough analysis yields selectivity estimations that can be used to optimize any combination of spatial and nonspatial selection and join operators. The accuracy of the formulae is evaluated through experimentation with various queries. In addition to their importance for spatial databases, the presented results can be applied in several other domains, where dependencies exist between operators.

Approximate spatio-temporal retrieval

Abstract: This paper proposes a framework for the handling of spatio-temporal queries with inexact matches, using the concept of relation similarity. We initially describe a binary string encoding for 1D relations that permits the automatic derivation of similarity measures. We then extend this model to various granularity levels and many dimensions, and show that reasoning on spatio-temporal structure is significantly facilitated in the new framework. Finally, we provide algorithms and optimization methods for four types of queries: (i) object retrieval based on some spatio-temporal relations with respect to a reference object, (ii) spatial joins, i.e., retrieval of object pairs that satisfy some input relation, (iii) structural queries, which retrieve configurations matching a particular spatio-temporal structure, and (iv) special cases of motion queries. Considering the current large availability of multidimensional data and the increasing need for flexible query-answering mechanisms, our techniques can be used as the core of spatio-temporal query processors.

Solving Non-binary CSPs Using the Hidden Variable Encoding

Abstract: Non-binary constraint satisfaction problems (CSPs) can be solved in two different ways. We can either translate the problem into an equivalent binary one and solve it using well-established binary CSP techniques or use extended versions of binary techniques directly on the non-binary problem. Recently, it has been shown that the hidden variable encoding is a promising method of translating non-binary CSPs into binary ones. In this paper we make a theoretical and empirical investigation of arc consistency and search algorithms for the hidden variable encoding. We analyze the potential benefits of applying arc consistency on the hidden encoding compared to generalized arc consistency on the non-binary representation. We also show that search algorithms for non-binary constraints can be emulated by corresponding binary algorithms that operate on the hidden variable encoding and only instantiate original variables. Empirical results on various implementations of such algorithms reveal that the hidden variable is competitive and in many cases better than the non-binary representation for certain classes of non-binary constraints.

Constraint-Based Processing of Multiway Spatial Joins

Abstract: A multiway spatial join combines information found in three or more spatial relations with respect to some spatial predicates. Motivated by their close correspondence with constraint satisfaction problems (CSPs), we show how multiway spatial joins can be processed by systematic search algorithms traditionally used for CSPs. This paper describes two different strategies, window reduction and synchronous traversal, that take advantage of underlying spatial indexes to prune the search space effectively. In addition, we provide cost models and optimization methods that combine the two strategies to compute more efficient execution plans. Finally, we evaluate the efficiency of the proposed techniques and the accuracy of the cost models through extensive experimentation with several query and data combinations.

Efficient image retrieval by exploiting vertical fragmentation

Abstract: In content-based retrieval systems, the goal of similarity search is to identify the k most similar images to a given example. Images are represented and queried by high-dimensional feature vectors encoding dominant characteristics like color and texture. We propose an efficient similarity search method that is robust to dimensionality and has optimal space complexity. Our approach fragments the feature vectors vertically, and computes the similarity of all images dimension by dimension. The innovation lies in gradually removing images that cannot participate in the response set. We show how to apply this scheme for two common similarity metrics, namely histogram intersection and euclidean distance. The implementation of our algorithm in Monet illustrates that core database technology supports image retrieval well, without special extensions. Finally, we report the effectiveness of our approach on real and synthetic data sets, and show significant improvements in response time yielded.