Tallis: A Statistical Approach for Dimension Reduction of Mixed-Type Variables

• Activity Number: 119 - Statistical Learning Applications for Autonomous Systems in Defense and National Security
• Type: Topic Contributed
• Date/Time: Monday, August 3, 2020 : 1:00 PM to 2:50 PM
• Sponsor: Section on Statistics in Defense and National Security
• Author(s): Alexander Foss*
• Companies: Sandia National Laboratories
• Keywords: Dimension reduction; mixed-type data; anomaly detection

Abstract

A common challenge in a broad array of disciplines is dimension reduction of variables of mixed-type. In this talk, a novel statistical model for dimension reduction of combinations of continuous, count, and categorical data is introduced and contrasted with existing approaches in the literature and practice. An Expectation-Maximization (EM) algorithm for estimating model parameters is derived, and an algorithm for parallel EM chains is discussed. The performance of the method is illustrated using examples drawn from anomaly detection and graph analysis applications.

SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525