Abstract for: Integrating Simulation, Machine Learning and High-performance Computing with EMEWS

In this presentation I will review ongoing efforts by our research group to facilitate the intersection of simulation, machine learning (ML) methods, and high-performance computing (HPC), three areas of continuing general interest and growth, to tackle the intricacies of complex systems modeling. I will provide an overview of how ML-driven HPC workflows can be leveraged to advance computational science and support decision making. In doing so, I will describe how our Extreme-scale Model Exploration with Swift (EMEWS) framework (https://emews.github.io) leverages advances in ML algorithms to enable large-scale model exploration of computational models, including natural history microsimulations and agent-based models, on HPC resources. I will demonstrate applications of our approach across scientific domains where the three pillars of simulation, ML and HPC provide the analytical platform for in silico experiments at the scales needed for deepening our understanding of important phenomena and to produce critical answers to pressing scientific and policy questions. Throughout, I will make the case for our overarching goal of improving interoperability, scalability, transparency and reproducibility in large-scale computational modeling.