Sparse Low-Rank Models for the Integration of Multiple Data Types

• Activity Number: 398
• Type: Invited
• Date/Time: Tuesday, August 6, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Eric Frazer Lock*+
• Companies: Duke University
• Keywords: Multi-block data ; Data integration ; Singular Value Decomposition ; Principal Component Analysis ; Multi-way data analysis

Abstract

Research in several fields now requires the analysis of datasets in which multiple high dimensional types of data are available for a common set of objects. We propose sparse linear decomposition methods for the integrated analysis of such datasets. These extend the Joint and Individual Variation Explained (JIVE) method, which gives a low-rank approximation capturing joint structure across data types, low-rank approximations for structured variation individual to each data type, and residual noise. The proposed approach is also closely related to sparse SVD, PCA, and tensor factorization methods. We describe an application to genomic data from multiple sources.