Expert-Guided Generative Topographic Modeling with Visual to Parametric Interaction

• Activity Number: 217
• Type: Invited
• Date/Time: Monday, August 5, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Graphics
• Author(s): Leanna House*+ and Chao Han and Scotland Charles Leman
• Companies: Virginia Tech and Virginia Tech and Virginia Tech
• Keywords: Visualization ; Human-Computer Interaction ; Clustering ; Text ; Generative Topographic Mapping

Abstract

Introduced by Bishop et al. in 1996, Generative Topographic Mapping (GTM) is a powerful nonlinear latent variable modeling approach for visualizing high-dimensional data. It has shown useful when typical linear methods fail. However, GTM still suffers from drawbacks. GTM's complex parameterization makes it hard to fit and sensitive to slight changes in the model and/or data. For this reason, we extend GTM to a visual analytics framework so that users may guide the parameterization and assess data from multiple GTM perspectives. Specifically, we develop the theory and methods for Visual to Parametric Interaction (V2PI) with data using GTM visualizations. The result is a dynamic version of GTM, V2PI-GTM, that fosters data exploration. We demonstrate the benefits of V2PI-GTM within the context of a text mining case study.