Identifiability Bounds for Recovering Governing Equations from Solution Data

arXiv·low signal

Pan and Bölcskei give theoretical conditions under which a ground-truth ODE can be uniquely and stably identified from multiple solution observations — a gap in scientific machine learning, where equation-discovery methods proliferate but rigorous identifiability guarantees do not. They provide quantitative identifiability bounds for both linear and nonlinear ODEs. Grounding for the reliability of data-driven dynamics discovery.