A New Approach to Fast Matrix Factorization for the Statistical Analysis of High-Dimensional Data (2011–2013)

Extracting interesting knowledge and information from experimental raw data sets and observations is pervasive throughout science, engineering, and medical science. This project will develop a novel and very fast approach to dimension reduction to enable the extraction of information from high-dimensional data sets. It uses matrix factorization to provide an approximation to the data matrix, whereby the essential features are captured by a smaller set of so-called metavariables. Statistical analyses can then be performed in terms of the metavariables to understand and draw inferences from complex data sets. Key applications include the discovery of new subclasses of cancer and their diagnosis and prognosis, as to be demonstrated.
Grant type:
ARC Discovery Projects
Funded by:
Australian Research Council