Lucent Technologies Inc.

International Business Machines Corporation

Kyuseok Shim

Rajeev R Rastogi

Rakesh Agrawal

Minos N Garofalakis

Sudipto Guha

Harpreet Singh Sawhney

King-Ip Lin

Ramakrishnan Srikant

Srinivasan Seshadri

Sridhar Ramaswamy

Aristides Gionis

Kaushik Chakrabarti

Apostol Ivanov Natsev

The present invention discloses a document descriptor extraction method and system. The document descriptor extraction method and system creates a document descriptor by generalizing input sequences within a document; factoring the input sequences and generalized input sequences; and selecting a document descriptor from the input sequences, generalized sequences, and factored sequences, preferably using minimum descriptor length (MDL) principles. Novel algorithms are employed to perform the generalizing, factoring, and selecting.

Document descriptor extraction method

A method for querying electronic data. The query method comprises creating wavelet-coefficient synopses of the electronic data and then querying the synopses in the wavelet-coefficient domain to obtain a wavelet-coefficient query result. The wavelet-coefficient query result is then rendered to provide an approximate result.

Approximate query processing using wavelets

Methods of imaging objects based on wavelet retrieval of scenes utilize wavelet transformation of plural defined regions of a query image. By increasing the granularity of the query image to greater than one region, accurate feature vectors are obtained that allow for robust extraction of corresponding regions from a database of target images. The methods further include the use of sliding windows to decompose the query and target images into regions, and the clustering of the regions utilizing a novel similarity metric that ensures robust image matching in low response times.

Methods of imaging based on wavelet retrieval of scenes

A new method for identifying a predetermined number of data points of interest in a large data set. The data points of interest are ranked in relation to the distance to their neighboring points. The method employs partition-based detection algorithms to partition the data points and then compute upper and lower bounds for each partition. These bounds are then used to eliminate those partitions that do contain the predetermined number of data points of interest. The data points of interest are then computed from the remaining partitions that were not eliminated. The present method eliminates a significant number of data points from consideration as the points of interest, thereby resulting in substantial savings in computational expense compared to conventional methods employed to identify such points.

Method for identifying outliers in large data sets

The present invention provides a method and system for sequential pattern mining with a given constraint. A Regular Expression (RE) is used for identifying the family of interesting frequent patterns. A family of methods that enforce the RE constraint to different degrees within the generating and pruning of candidate patterns during the mining process is utilized. This is accomplished by employing different relaxations of the RE constraint in the mining loop. Those sequences which satisfy the given constraint are thus identified most expeditiously.

System and method for constraint based sequential pattern mining

A method of data classification using a decision tree having nodes is disclosed, along with an apparatus for perming the method. Periodically or after a certain number of nodes of the tree are split, the partially built tree is pruned. During the building phase the minimum cost of subtrees rooted at leaf nodes that can still be expanded (&ldquo;yet to be expanded nodes&rdquo;)is computed. With the computation of the minimum subtree cost at nodes, the nodes pruned are a subset of those that would have been pruned anyway during the pruning phase, and they are pruned while the tree is still being built.

Decision tree classifier with integrated building and pruning phases

An electronic data mining process for mining from an electronic data base using an electronic digital computer a listing of commercially useful information of the type known in the art as an association rule containing at least one uninstantiated condition. For example, the commercially useful information may be information useful for sales promotion, such as promotion of telephone usage. The computer retrieves from the database a plurality of stored parameters from which measures of the uninstatiated condition can be determined. The computer uses a dynamic programming algorithm and iterates over intervals or sub-ranges of the parameters to obtain what is called an at least partially optimized association rule, as optimized intervals or sub-ranges of at least some of the retrieved parameters, for example, time intervals of high usage of certain types of telephone connections. These optimized intervals are provided as the listed commercially useful information. The amount of needed iteration is reduced in some cases by using so-called bucketing and divide-and-conquer techniques. Extension of the process for a plurality of uninstantiated conditions is described.

Method for mining association rules in data

The present invention relates to a computer method, apparatus and programmed medium for clustering large databases. The present invention represents each cluster to be merged by a constant number of well scattered points that capture the shape and extent of the cluster. The chosen scattered points are shrunk towards the mean of the cluster by a shrinking fraction to form a representative set of data points that efficiently represent the cluster. The clusters with the closest pair of representative points are merged to form a new cluster. The use of an efficient representation of the clusters allows the present invention to obtain improved clustering while efficiently eliminating outliers.

Programmed medium for clustering large databases

The present invention relates to a computer method, apparatus and programmed medium for clustering databases containing data with categorical attributes. The present invention assigns a pair of points to be neighbors if their similarity exceeds a certain threshold. The similarity value for pairs of points can be based on non-metric information. The present invention determines a total number of links between each cluster and every other cluster bases upon the neighbors of the clusters. A goodness measure between each cluster and every other cluster based upon the total number of links between each cluster and every other cluster and the total number of points within each cluster and every other cluster is then calculated. The present invention merges the two clusters with the best goodness measure. Thus, clustering is performed accurately and efficiently by merging data based on the amount of links between the data to be clustered.

Method, apparatus and programmed medium for clustering databases with

A method and system are disclosed for performing spatial similarity joins on high-dimensional points that represent data objects of a database. The method comprises the steps of: generating a data structure based on the similarity distance .epsilon. for organizing the high-dimensional points, traversing the data structure to select pairs of leaf nodes from which the high-dimensional points are joined, and joining the points from selected pairs of nodes according to a joining condition based on the similarity distance .epsilon.. An efficient data structure referred to as an .epsilon.-K-D-B tree is disclosed to provide fast access to the high-dimensional points and to minimize system storage requirements. The invention provides algorithms for generating the .epsilon.-K-D-B tree using biased splitting to minimize the number of nodes to be examined during join operations. The traversing step includes joining selected pairs of nodes and also self-joining selected nodes. Alternatively, the data structure is an R+tree generated using biased splitting.

Method and system for performing spatial similarity joins on

In a data processing system, association rules are used to determine correlations of attributes of collected data, thereby extracting insightful information therefrom. In solving an optimized association rule problem where multiple instantiations for at least one uninstantiated attribute are required, unlike prior art, not all possible instantiations are considered to realize an optimized set of instantiations. Rather, using inventive pruning techniques, only selected instantiations need to be considered to realize same. In accordance with the invention, instantiations are assigned weights and are subject to pruning in an order dependent upon their weight. The weighted instantiations are tested based on selected criteria to identify, for example, those instantiations, consideration of which for the optimized set would be redundant in view of other instantiations to be considered. The identified instantiations are disregarded to increase the efficiency of determining the optimized set.

Technique for effectively instantiating attributes in association rules

A system and method for discovering similar time sequences in a database of time sequences includes a computer-implemented program which first breaks each sequence into small windows. The windows from the first sequence are compared to selected windows from the second sequence to determine which windows are similar. Pairs of similar windows are then stitched together when certain stitching constraints are met to establish pairs of similar subsequences. Likewise, pairs of similar subsequences are stitched together, and the lengths of the stitched subsequences are then compared to the overall length of the time sequences to determine whether the time sequences meet a similarity criteria.

System and method for discovering similar time sequences in databases

An application program is tightly coupled to a relational database without changing the software of the database management system. A first user defined function (udf) allocates a work area in the address space of the database management system. Then, a second udf embodying the computational steps of the application program undertakes the computational steps within the dbms work area and saves the results in the work area. Importantly, the second udf does not return an answer to the application program during the processing of records, thereby eliminating copying costs and context switch costs. When processing is complete, the results are transmitted to the application program and the work area is released.

Growing community of inventors

Bedminster, NJ, United States of America

Kyuseok Shim