Konig A., Weikum G.: Combining Histograms and Parametric Curve Fitting for Feedback-Driven Query Result-Size Estimation. VLDB Conf. (1999) 423-434  Home/Search   Document Details and Download   Summary   Related Articles   Check

....centralized environments to distributed environments. The problem here is to identify the independently operated databases that are likely to contain the top N tuples for a given query; such a problem does not exist in a centralized environment. 2. Some recent techniques to construct histograms ([5, 9, 11, 13, 18, 19]) are employed here, although there is a signi cant di erence. Histograms were traditionally used to estimate the number of tuples satisfying a certain query condition. In this paper, we modify existing techniques and propose a new technique so that they can be used to estimate the distance of ....

....second point is the corresponding mileage value. 3) Since there may be too many intervals, requiring excessive storage, adjacent intervals are merged until a xed number of, say r, intervals remain. The criterion to choose which adjacent intervals to merge is given by the Greedy Merge algorithm [13]. For every two adjacent intervals, the combined interval contains a straight line with a least square approximation to the tuples within it. The Greedy Merge algorithm chooses the combined interval with the smallest least square error. This process of combining adjacent intervals continues until ....

A. Konig and G. Weikum. Combining Histograms and Parametric Curve Fitting for FeedbackDriven Query Result-Size Estimation. VLDB Conference, 1999.

....centralized environments to distributed environments. The problem here is to identify the independently operated databases that are likely to contain the top N tuples for a given query; such a problem does not exist in a centralized environment. 2. Some recent techniques to construct histograms ([5, 9, 11, 13, 18, 19]) are employed here, although there is a signi cant di erence. Histograms were traditionally used to estimate the number of tuples satisfying a certain query condition. In this paper, we modify existing techniques and propose a new technique so that they can be used to estimate the distance of ....

....of the two end points of a line segment. 3) Since there may be too many intervals, requiring excessive storage, adjacent intervals are merged until a xed number of, say r, intervals remain. The criterion to choose which adjacent intervals to merge is given by the Greedy Merge algorithm [13]. For every two adjacent intervals, the combined interval contains a straight line with a least square approximation to the tuples within it. The Greedy Merge algorithm chooses the combined interval with the smallest least square error. This process of combining adjacent intervals continues until ....

A. Konig and G. Weikum. Combining Histograms and Parametric Curve Fitting for FeedbackDriven Query Result-Size Estimation. VLDB Conference, 1999.

....approach that we follow in [5] and also in this paper. Multidimensional density estimation is an active research field. The main techniques comprise the use of multidimensional histograms [13] Some variations over histograms include the use of parametric curve fitting techniques inside buckets [10], self tuning histograms [1] and lately, multidimensional histograms for dealing with real valued attributes [9] Other multidimensional density estimation techniques are wavelets [12] and fractal dimension concepts [8, 2] 7 Conclusions In this paper, we have presented a new robust scheme for ....

A. C. Konig and G. Weikmun. Combining histograms and parametric curve fitting for feedback-driven query result-size estimation. In Proceedings of the Twenty-fifth International Conference on Very Large Databases (VLDB'99), Sept. 1999.

....1000 . Table 1: A Histogram for Car Mileages From the histogram in Table 1, it can be seen that if there are 10,000 tuples in this relation, then the probability that a tuple with mileage in the range [0,10k) is 200 10000 = 0.02. b) Greedy Merge Method This method was proposed in [10]. Initially, the range of values is partitioned into a large number of subranges of equal width. As in the simple interval construction method, the number of tuples which have values within each subrange is counted. Associ4 ated with each subrange, an error of estimation can be computed. For a ....

....This is repeated until a certain number of subranges is reached. At that point, the counts for the different subranges are kept. If proper statistics are kept for each subrange, then determining which adjacent subranges to be merged can be carried out efficiently. The details can be found in [10]. 3.2 Criterion for Selecting Databases Optimally Definition 1 Suppose a user is interested in retrieving the k nearest neighbors to a submitted query Q. Databases fD i ; 1 i ng are optimally ranked in the order D 1 ; D 2 : Dn , if for every k, there exists a t such that D 1 ; D 2 ; D ....

[Article contains additional citation context not shown here]

A. Konig and G. Weikum. Combining Histograms and Parametric Curve Fitting for FeedbackDriven Query Result-Size Estimation, VLDB, 1999.

.... the milestone uniformity and independence assumptions [SAC 79] Although simple to use in a query optimizer, these assumptions are pessimistic and unrealistic [CHR 84] Modern methods include histograms [POO97] kernel estimators [BKS99] wavelets [VW99] and hybrid methods using query feedback [KW 99] Methods for selectivity estimation of range queries in spatial datasets use multidimensional histograms [TS 96] or arguments from the theory of fractals [BF 95] It should be noted that most of these methods are susceptible to the dimensionality curse [SIL 96] SCO 92] Analytical estimates ....

A. Christian Kvnig, G. Weikum - "Combining Histograms and Parametric Curve Fitting for Feedback-Driven Query Result-size Estimation".VLDB 1999, pp.423-434.

.... site) as a basis for planning global queries in a Web mediator [27] 28] This broad importance of statistics management has led to a plethora of approximation techniques, for which [15] have coined the general term data synopses : advanced forms of histograms [30, 16, 20] spline synopses [22, 23], sampling [6, 17, 14] and parametric curve fitting techniques [34, 9] all the way to highly sophisticated methods based on kernel estimators [2] or Wavelets and other transforms [26, 25, 4] However, most of these techniques take the local viewpoint of optimizing the approximation error for a ....

....2 : SELECT R 1 .A 1 FROM R 1 , R 2 , R 3 WHERE R 1 .A 2 = R 2 .A 3 AND R 2 .A 4 = R 3 .A 5 Here we introduce R # : R 1 ## R 2 ## R 3 . Then Min(q 2 ) # .A 1 . 2. 3 Spline Synopses As the underlying statistics representation, we use spline synopses, which are described in detail in [22, 23]. Our results also apply (with some adaptation) to other data reduction techniques (e.g. histograms) Spline synopses have particular properties that are advantageous in our physical design context. The approximation of a distribution is again a data distribution T , i.e. for every ....

A. Konig and G. Weikum. Combining Histograms and Parametric Curve Fitting for Feedback-Driven Query Result-size Estimation. In 25th International Conference on Very Large Databases, 1999.

No context found.

Konig A., Weikum G.: Combining Histograms and Parametric Curve Fitting for Feedback-Driven Query Result-Size Estimation. VLDB Conf. (1999) 423-434

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC