Methods of decomposing complex data

Abstract

The invention includes a computer implemented process to identify at least one pattern and its distribution in a set of data for the purpose of interpreting the data. The process comprises (a) representing a set of data by an original data matrix D residing in a storage device, and; (b) decomposing the set of data into a set of patterns represented by a matrix F and their distribution represented by a matrix A, wherein the matrix F represents the set of patterns needed to describe the data and the matrix A represents the distribution of the set of patterns within the data matrix D, the decomposing comprising performing a Bayesian-based Monte Carlo calculation using at least the data matrix D to determine the matrices A and F, wherein the matrices A and F reconstruct the data matrix D and are more amenable to analysis than the data matrix D. Application of the process to environmental, biological and medical, econometric, and other fields is included in the invention.

Dependencies are inferred from the claim text.