Online Program Home
My Program

Abstract Details

Activity Number: 461 - Bayesian Statistical Methods for High-Throughput Toxicity Testing and Risk Assessment
Type: Topic Contributed
Date/Time: Wednesday, July 31, 2019 : 8:30 AM to 10:20 AM
Sponsor: Section on Risk Analysis
Abstract #307341 Presentation
Title: Linked Matrix Factorization
Author(s): Michael O'Connell*
Companies: Miami University
Keywords: dimension reduction; multi-source data; high-dimensional data

There have been several methods devoloped in recent years for decompositions of multisource data, consisting of more than one matrix with at least one shared dimension. Several methods have been developed for the simultaneous dimension reduction and decomposition of multiple matrices. Typically these methods assume that either features are shared for different sample sets (horizontal integration) or that samples are shared for different feature sets (vertical integration). However, these algorithms do not allow for simultaneous horizontal and vertical integration. For data sets that have shared sample sets and shared feature sets, we developed the Linked Matrix Factorization algorithm (LMF), an alternating least squares-based method that allows for decomposition of three matrices simultaneously when one matrix shares its sample set with one matrx and its feature set with another. We illustrate the application of LMF using a toxicology data set. In this data set, the toxicity matrix shares its sample set (cell lines) with a gene expression matrix, and it shares its feature set (chemicals) with a chemical attribute matrix.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2019 program