Speaker: 李静怡, 北京师范大学

Inviter: 熊世峰

Title: Bayesian inference for a Discrepancy Decomposition Model in Calibration Experiments

Language: Chinese

Time & Venue: 2026.03.17 09:30-10:30 思源楼 803

Abstract: Computer simulation models are widely used to study complex physical systems. A related topic is the calibration problem, which aims at learning about the values of physical parameters in the model based on observations. To understand the true underlying physical system, we need to effectively estimate such parameters. However, existing calibration methods have limitations in addressing this issue due to model identifiability. This paper proposes a method based on the discrepancy decomposition model, to describe the discrepancy between the physical system and the computer model. The proposed model possesses a clear interpretation, and more importantly, it is identifiable under mild conditions. Under this model, we present Baysian estimators of the physical parameters and the discrepancy functions. Numerical examples show that the proposed method is capable of accurately estimating the physical parameters and quite robust to model assumptions.