International Business Machines Corporation

At&t Intellectual Property I, L.p.

At+t Corp.

Michail Vlachos

Philip Shilung Yu

Abderrahim Labbi

Philip Shi-lung Yu

Vittorio R Castelli

Francesco Fusco

Vasileios Vasileiadis

Reinhard Wolfram Heckel

Charalampos Pozidis

Divesh Srivastava

Kun-Lung Wu

Thomas P Parnell

Marc Philippe Stoecklin

Nikolaos Koudas

Shyh-Kwei Chen

Celestine Duenner

Sudipto Guha

Dimitrios Gunopulos

Anastasios Zouzias

Thomas Mittelholzer

Olivier Verscheure

Eric P Bouillet

Andreas Kind

Lisa Amini

Kubilay Atasu

Christos Boutsidis

Nikolaos Freris

Deepak Srinivac Turaga

Deepak Srinivao Turaga

A method for parallelizing a training of a model using a matrix-factorization-based collaborative filtering algorithm may be provided. The model can be used in a recommender system for a plurality of users and a plurality of items. The method includes providing a sparse training data matrix, selecting a number of user-item co-clusters, and building a user model data matrix by matrix factorization such that a computational load for executing the determining updated elements of the factorized sparse training data matrix is evenly distributed across the heterogeneous computing resources.

Load-balancing training of recommender system for heterogeneous systems

A method for data mining on compressed data vectors by a certain metric being expressible as a function of the Euclidean distance is suggested. In a first step, for each compressed data vector, positions and values of such coefficients having the largest energy in the compressed data vector are stored. In a second step, for each compressed data vector, the coefficients having not the largest energy in the compressed data vector are discarded. In a third step, for each compressed data vector, a compression error is determined in dependence on the discarded coefficients in the compressed data vector. In a fourth step, at least one of an upper and a lower bound for the certain metric is retrieved in dependence on the stored positions and the stored values of the coefficients having the largest energy and the determined compression errors.

Method and device for data mining on compressed data vectors

A method for guided keyword-based exploration of data stored in a database includes providing, with a processing device, valid query templates for the data, wherein the provided valid query templates include static parts and dynamic parts; selecting those of the provided valid query templates that match a user-provided keyword; generating valid queries from the selected valid query templates using the data; and querying the data using a user-selected valid query selected from the generated valid queries.

Guided keyword-based exploration of data

Methods and apparatus are provided to determine entities and attributes dependencies for creating recommendations of items or entities using a highly scalable architecture. For example, a user may be recommended an item if a probability model of the method determines that the user relates to the item although the user has no contact to the item before the method is performed. The methods and apparatus provide a data structure representing a matrix having rows representing entities and columns representing attributes of the entities. Each entity of the entities of the data structure may include a user and each attribute of the attributes of the data structure may include an item. A cell of the matrix may be formed by a component pair including an entity and an attribute. In this manner, the methods and apparatus provide an efficient way for processing the probability model.

System and method for a scalable recommender system using massively parallel processors

Embodiments of the present invention disclose a method, computer program product, and system for data obfuscation and right-protection. An initial matrix X, represents the initial data set of the application and final matrix Xis obtained from X. The final matrix Xis obtained by performing one of the following operations X=(P(X)+E)F; X=P(X)F+E; and X=P(XF)+E. Where P(.) is a projection operator that projects an input initial matrix in a space having a lower dimension than the input matrix, E represents a noise matrix, and F represents a matrix as a perturbation series. The matrix F is represented as a perturbation series, whose leading term is the identity matrix I, one or more higher-order terms of the perturbation series embedding a secret, multiplicative noise, so as for a matrix multiplied by the matrix F is right-protected.

Obfuscation and protection of data rights

A computer-implemented method for identifying relationships between entities includes accessing a first data structure being a two-dimensional array of scalar elements (e, e, e) representable as a matrix, each of the scalar elements capturing a relationship between two entities; reorganizing the first data structure by clustering the scalar elements separately on each dimension of the two-dimensional array, to obtain a second data structure, representable as a K&#xd7;M block matrix, wherein each block is a reordered sequence of rows and/or columns of the first data structure; compacting the second data structure by: determining two parallel block sequences, which are the most similar according to a given distance measure, the parallel block sequences being either distinct rows or distinct columns of blocks of the second data structure; and reorganizing the second data structure by merging the two determined sequences into a single block sequence.

Identifying relationships between entities using two-dimensional arrays of scalar elements and a block matrix and displaying dense blocks

Embodiments include processing a data structure representing a dependency matrix having columns representing respective first components and rows representing respective second components. Aspects include assigning each cell of the matrix a value indicative of the level of dependency or indicative of an unknown dependency of a pair of first and second components forming the cell and assigning each component of the first and second components an affiliation vector indicative of the strength of affiliation of the component to N predefined initial clusters of cells of the matrix. Aspects also include determining a probability model using the affiliations vectors parameters and estimating the parameters of the probability model for a plurality of different numbers of clusters starting from the initial number N of clusters. Aspects further include computing a score for the parameters of the probability model estimated and selecting the parameters of the probability model with the highest computed score.

Method and system for identifying dependent components

Identifying relationships between entities using two-dimensional array of scalar elements, and a block matrix

Exemplary embodiments include a method for re-ordering and visualizing a matrix in the presence of a data hierarchy stored on a computer system, the method including receiving a matrix, in the computer system, the matrix having rows and columns, reordering the matrix so that groups of related rows and columns are brought in adjacent matrix positions, wherein reordering the matrix A obeys constraints imposed by the data hierarchy on at least one of the rows and the columns and reordering the hierarchy on at least one of the rows and columns, wherein affinities in the data hierarchy are extracted by the reordering of the data hierarchy.

Matrix re-ordering and visualization in the presence of data hierarchies

Systems and methods for embedding metadata such as personal patient information within actual medical data signals obtained from a patient are provided wherein two watermarks, a robust watermark and a fragile watermark are embedded in a given medical data signal. The robust watermark includes a binary coded representation of the metadata that is incorporated into the frequency domain of the medical data signal using discrete Fourier transformations and additive embedding. Error correcting code can also be added to the binary representation of the metadata using Hamming coding. A given robust watermark can be incorporated multiple times in the medical data signal. The fragile watermark is added on top of the modified medical signal containing the robust watermark in the spatial domain of the modified medical signal. The fragile watermark utilizes hash function to generate random sequences that are incorporated through the medical data signal.

Systems and methods for metadata embedding in streaming medical data

Methods and a device for providing a compressed index of binary records. A method includes: sorting the records by content of a predetermined field of the record, indexing the field from one of the records in a line of a bitmap index, compressing bits in a column of the bitmap index by replacing a group of successive bits with a code, where the sorting includes the steps of assigning, for each record, a hash bucket of a hash table on a basis of a locality sensitive hash function on the contents of the predetermined field, so that the probability for two of the records to be assigned to the same has bucket increases with the similarity of the contents of the predetermined field between the records, and where at least one step of the computer implemented method is executed on a computer device.

Network analysis

A system and method for rights protection of a dataset that includes multiple trajectory objects includes determining an intensity power for embedding a watermarking key in a data trajectory. The data trajectory is modified to embed a watermarking key at the intensity power such that the intensity power guarantees an original pair-wise relationship between distance-based neighboring objects before and after embedding of the key such that a modified trajectory provides a watermarked version of the data trajectory.

Systems and methods for rights protection of datasets with dataset structure preservation

A method for providing a compressed index for a stream of binary data records comprises steps of indexing a field from each record in a bitmap index, compressing stored bits in each column of the bitmap index by replacing a group of successive bits with a code and outputting the code. There is provided at least one of a first code for replacing a sequence of a first filling, a literal and a second filling, and a second code for replacing a sequence of a first literal, a filling and a second literal. In this context, a filling is a sequence of bits with the same value and a literal is a sequence of bits with different values.

Growing community of inventors

Rueschlikon, Switzerland

Michail Vlachos