Invited speaker

Rose Yu — UC San Diego

Short bio

Rose Yu is assistant professor at UC San Diego department of Computer Science and Engineering and Halıcıoğlu Data Science Institute. Her research interests lie primarily in machine learning, especially for large-scale spatiotemporal data. She is generally interested in deep learning, optimization, and spatiotemporal reasoning. Her work has been applied to learning dynamical systems in sustainability, health and physical sciences.

Deep Generative Models for Missing Data in Temporal Sequences

Missing data imputation is a fundamental problem in spatiotemporal modeling, from motion tracking, video analysis, to the dynamics of physical systems. Traditional methods suffer from error propagation which becomes catastrophic for imputing long-range sequences. In this talk, I will introduce deep generative models for sequence imputation. Specifically, I will describe (1) Non-AutOregressive Multiresolution Imputation (NAOMI): a novel deep generative model to impute long-range trajectories given arbitrary missing patterns. and (2) Disentangled Imputed Video autoEncoder (DIVE): a deep generative model that imputes and predicts future video frames in the presence of missing data. I will showcase these models on benchmark time series and video imputation tasks.

Comments are closed.