US 12,602,449 B2 · Pattern change discovery between high dimensional data sets

Abstract

The general problem of pattern change discovery between high-dimensional data sets is addressed by considering the notion of the principal angles between the subspaces is introduced to measure the subspace difference between two high-dimensional data sets. Current methods either mainly focus on magnitude change detection of low-dimensional data sets or are under supervised frameworks. Principal angles bear a property to isolate subspace change from the magnitude change. To address the challenge of directly computing the principal angles, matrix factorization is used to serve as a statistical framework and develop the principle of the dominant subspace mapping to transfer the principal angle based detection to a matrix factorization problem. Matrix factorization can be naturally embedded into the likelihood ratio test based on the linear models. The method may be unsupervised and addresses the statistical significance of the pattern changes between high-dimensional data sets.

Claims18

Dependent claims

A method for distinguishing statistical significance of changes between first semantic data representing at least one of multimedia data, social network data, and news data at a first time and second semantic data representing at least one of multimedia data, social network data, and news data at a second time, the second semantic data being different from the first semantic data, comprising:

linearly modelling at least a dominant subspace of the first semantic data and the second semantic data in a linear model defining a respective matrix;
factoring each matrix into a feature basis matrix and an indicator matrix;
determining principal angles which are independent of respective magnitudes of the changes, dependent on at least the feature basis matrices and the indicator matrices;
adaptively determining a threshold of statistical significance based on the linear model;
determining a statistical significance of the changes based on at least the principal angles and the threshold;
and
selectively outputting a subset of at least one of the first semantic data, and the second semantic data through a computer communication network interface associated with a prediction of a change in trajectory of the at least one of multimedia data, social network data, and news data, dependent on the statistical significance of the changes.

4 more elements

A system for determining a statistical significance of changes over time in a semantic data set representing at least one of multimedia data, social network data, and news data, having a dimensionality scale which is of at least thousands, comprising:

a memory configured store the changes over time in the semantic data set;
and
at least one automated processor configured to:
linearly model at least a dominant subspace of the semantic data set, to produce respective linear models of the semantic data set at different times, each comprising a respective matrix;
factor each respective matrix into a respective feature basis matrix and a respective indicator matrix;
determine a set of principal angles between the linear model of the semantic data set at the different times, dependent on at least the respective feature basis matrices and the respective indicator matrices;
and
adaptively determine a threshold of statistical significance based on the linear model;
and
determine a statistical significance of changes over time based on an analysis of at least the principal angles;
and
a computer communication network interface configured to communicate a subset of the semantic data set associated with a prediction of a change in trajectory of the at least one of multimedia data, social network data, and news data, dependent on the determined statistical significance of the changes.

9 more elements

Dependencies are inferred from the claim text.

Pattern change discovery between high dimensional data sets

Life of this patent

Pattern change discovery between high dimensional data sets

Life of this patent

Abstract

Claims18