Online Program Home
My Program

Abstract Details

Activity Number: 177 - Section on Statistical Learning and Data Science CPapers 2
Type: Contributed
Date/Time: Monday, July 30, 2018 : 10:30 AM to 12:20 PM
Sponsor: Section on Statistical Learning and Data Science
Abstract #329122
Title: Semi-Orthogonal Matrix Factorization
Author(s): Yutong Li* and Ruoqing Zhu and Annie Qu
Companies: University of Illinois at Urbana-Champaign and University of Illinois Urbana-Champaign and University of Illinois at Urbana-Champaign
Keywords: Non-negative Matrix Factorization; Orthogonal Constraint; Binary Matrix Factorization; Dimension Reduction; Cayley Transformation

Non-negative Matrix Factorization(NMF) is a popular clustering and dimension reduction method by decomposing a non-negative matrix into the product of two lower dimension matrices composed of basis vectors. In this paper, we propose a Semi-orthogonal NMF method that enforces one of the matrix to be orthogonal, thereby preserving the rank of the factorization and providing an alternative interpretation. Our method guarantees strict orthogonality by implementing the Cayley Transformation to force the solution path to be on the Stiefeld Manifold, as opposed to the approximated orthogonality solutions in existing literature. We apply a line search update scheme along with an SVD-based initialization which results in rapid convergence of the algorithm compared to other popular update schemes. In addition, we present two separate formulations of our method for both continuous and binary design matrices. Through experiments on various simulated data sets, we show that our model has an advantage over other NMF variations regarding the accuracy of the factorization, rate of convergence, and especially the degree of orthogonality, while also being computationally competitive.

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

Back to the full JSM 2018 program