Physics-Informed Gaussian Process Learning on Lie Groups

What is it about?

Gaussian process learning has shown remarkable success in learning and predicting dynamics by using only measured data in Euclidean spaces. However, many mechanical systems are actually living on Lie groups, which means that the traditional method does not work any more. We encode the physical prior knowledge into the learning process to construct physics-informed Gaussian process learning algorithms for mechanical systems on Lie groups, and also for a certain class of systems on homogeneous manifolds.

Featured Image