Seminar on Mechanism-Fused Statistical Inference

Abstract

This seminar explores the fusion of mechanistic modeling and statistical inference, presenting a range of methodologies for parameter estimation, curve fitting, and uncertainty quantification. Beginning with an introduction to mechanistic models, it covers nonlinear least squares, trajectory matching, and nonparametric curve fitting with roughness penalties. Advanced topics include gradient matching, generalized profiled estimation, and local polynomial methods. We further examine equation matching techniques and sparse additive models for ordinary differential equations (ODEs). The seminar also delves into Gaussian processes and their applications in ODE modeling, culminating in a discussion on the novel use of physics-informed neural networks (PINNs) for analyzing Argo data.

Time and Location

2022 Autumn, USTC Management Science Building and Online

Resources

Our materials are stored in Baidu Cloud Drive: PPT.