Probabilistic Numerics for ODEs 8: ODE filters can efficiently solve Inverse Problems

This video is part of a ten-part spotlight series on Probabilistic Numerical Methods for (ordinary) differential equations.

In this eighth video, Emilia Magnani presents work by Kersting et al. that shows probabilistic ODE filters provide differentials of the solution with respect to the vector field at essentially no cost (even less than explicit automatic differentiation itself), and thus allow efficient inference in inverse problems. (

The probnum package is under active development at

Source of this “Tübingen Machine Learning” AI Video

AI video(s) you might be interested in …