S6A: “Small Data” Approaches in Earthquake Engineering

2021 EERI Annual Meeting:
This session seeks to investigate the opportunities afforded by combining physics-based models with data-based technologies like machine learning to address challenges in earthquake risk assessment and performance-based engineering. Availability of large amounts of data to guide predictive models is causing a renaissance in many aspects of human endeavor including computer science, medicine and engineering.

The presentations in this session will describe several ways in which scientific knowledge can be combined with machine learning methods using case studies of on-going research illustrating multiple research themes in science-guided machine learning. Moreover, the speakers will discuss a reliability-based approach to evaluating the practical benefit (or lack thereof) provided by machine learning models in terms of reduced epistemic uncertainty in seismic risk estimation and will present a “small data” approach for efficient hazard-consistent collapse risk estimation.