Computational Mathematics Colloquium | Inference of Dynamic Systems from Noisy and Sparse Data via Manifold-constrained Gaussian Processes

Thursday, January 14, 2021 3:30 pm - 3:30 pm EST (GMT -05:00)

Speaker

Samuel Wong | Department of Statistics and Actuarial Science, University of Waterloo
https://swong.ca/index.html

Title

Inference of Dynamic Systems from Noisy and Sparse Data via Manifold-constrained Gaussian Processes 

Abstract

Ordinary differential equations are a ubiquitous tool for modeling behaviors in science, such as gene regulation, epidemics and ecology. An important problem is to infer and characterize the uncertainty of parameters that govern the equations. In this talk I will present an accurate and fast inference method using manifold-constrained Gaussian processes, such that the derivatives of the Gaussian process must satisfy the dynamics of the differential equations. Our method completely avoids the use of numerical integration and is thus fast to compute. Our construction is embedded in a principled statistical framework and is demonstrated to yield fast and reliable inference in a variety of practical problems, including when a system component is unobserved.

This is joint work with Shihao Yang (Georgia Tech) and Samuel Kou (Harvard).