Discrepancy Modeling with Physics Informed Machine Learning

This video describes how to combine machine learning with classical physics models to correct for discrepancies in the data (e.g., from nonlinear friction, wind resistance, etc.). Several examples are covered, from modern robotics, to classical connections with Galileo v. Aristotle, and Kepler v. Ptolemy. The examples in this video highlight work and discussions with Prof. Nathan Kutz, especially connections to classical scientific discoveries.

Citable link for this video at: https://doi.org/10.52843/cassyni.ftzlk9
@eigensteve on Twitter
eigensteve.com
databookuw.com

Papers discussed within:
https://arxiv.org/abs/1909.08574 [Double Pendulum]
https://www.frontiersin.org/articles/10.3389/frai.2020.00025/full [Ball Drop]

This video was produced at the University of Washington

%%% CHAPTERS %%%
0:00 Introduction
2:27 Double Pendulum Experiment (Example)
4:28 Hybrid Physics + Machine Learning Models
8:32 Analogy with Planetary Motion
10:37 Galileo’s Ball Drop Experiment