Learning from Heterogeneous Data with Extended Variational Autoencoder

2102 

Contributed Abstract 

Poster 

Jaeyoung Lee (1), Hongxiao Zhu (1)

(1) Virginia Tech, N/A

Hongxiao Zhu  
Virginia Tech

Jaeyoung Lee  
Virginia Tech

Jaeyoung Lee  
Virginia Tech

Large-scale datasets, such as images and texts, often exhibit complex heterogeneous structures caused by diverse data sources, complex experimental designs, or latent subpopulations. Supervised learning from such data is challenging, as it requires capturing relevant information from ultra-high-dimensional data while accounting for structural heterogeneity. We propose a unified framework that addresses both challenges simultaneously, facilitating effective feature extraction, structural learning, and robust prediction. The proposed framework employs an extended Variational Autoencoder for learning and prediction. Specifically, two types of latent variables are learned via the Variational Autoencoder: the low-dimensional latent features and a latent stick-breaking process that characterizes the clustering structure of the samples. The latent features serve as predictors for the response and the latent stick-breaking process serves as a "gating function" for mixture-of-experts prediction. This generalized framework reduces to a supervised Variational Autoencoder when the number of cluster is one, and reduces to a stick-breaking Variational Autoencoder when the latent features are

Variational Autoencoder|Data Heterogeneity |Stick-breaking Process|Supervised machine learning| |

Biometrics Section

High Dimensional Regression