A Deep Bayesian Active Learning Framework for Temporal Multimodal Data

Priyadip Ray | 19-ERD-009

Executive Summary

We are developing a high-performance, active-learning system for dynamic modeling that is capable of combining nonstatic data from disparate sources and learning the underlying temporal structure of that data in order to predict outcomes. This research has broad application in fields such as healthcare, energy, and nuclear nonproliferation.

Publications, Presentations, and Patents

Nguyen, S., et al. 2019. "Task State Classification of Four-Dimensional fMRI Time Series Using Deep Residual Convolutional Neural Networks." International Conference on Medical Imaging and Case Reports, Boston, MA, November 2019. LLNL-POST-796919

Ray, P. 2020a. "Estimating Vector Fields from Noisy Time Series." Learning for Dynamics & Control (L4DC). LLNL-CONF-800042