IEEE VIS 2024 Content: SurroFlow: A Flow-Based Surrogate Model for Parameter Space Exploration and Uncertainty Quantification

SurroFlow: A Flow-Based Surrogate Model for Parameter Space Exploration and Uncertainty Quantification

JINGYI SHEN - The Ohio State University, Columbus, United States. The Ohio State University, Columbus, United States

Yuhan Duan - The Ohio State University, Columbus, United States. The Ohio State University, Columbus, United States

Han-Wei Shen - The Ohio State University , Columbus , United States. The Ohio State University , Columbus , United States

Download preprint PDF
Download camera-ready PDF

Room: Bayshore I

2024-10-16T13:18:00ZGMT-0600Change your timezone on the schedule page
2024-10-16T13:18:00Z
Exemplar figure, described by caption below
In our work, we introduce SurroFlow, a novel normalizing flow-based surrogate model, to learn the invertible transformation between simulation parameters and simulation outputs. The model not only allows accurate predictions of simulation outcomes for a given simulation parameter but also supports uncertainty quantification in the data generation process. Additionally, it enables reverse prediction of simulation parameters of a given simulation data. We integrate SurroFlow and a genetic algorithm as the backend of a visual interface to support effective user-guided ensemble simulation exploration and visualization.
Fast forward
Keywords

Surrogate model, normalizing flow, uncertainty quantification, parameter space exploration

Abstract

Existing deep learning-based surrogate models facilitate efficient data generation, but fall short in uncertainty quantification, efficient parameter space exploration, and reverse prediction. In our work, we introduce SurroFlow, a novel normalizing flow-based surrogate model, to learn the invertible transformation between simulation parameters and simulation outputs. The model not only allows accurate predictions of simulation outcomes for a given simulation parameter but also supports uncertainty quantification in the data generation process. Additionally, it enables efficient simulation parameter recommendation and exploration. We integrate SurroFlow and a genetic algorithm as the backend of a visual interface to support effective user-guided ensemble simulation exploration and visualization. Our framework significantly reduces the computational costs while enhancing the reliability and exploration capabilities of scientific surrogate models.