Data-driven discovery of Green's functions

TL;DR

This paper develops deep learning algorithms and theoretical insights to discover Green's functions of linear PDEs from data, connecting PDE learning with numerical linear algebra and rational neural networks.

Contribution

It introduces a rigorous algorithm for learning Green's functions, extends randomized SVD to Hilbert--Schmidt operators, and proposes rational neural networks for approximating singular functions.

Findings

Derived a learning rate for Green's functions of elliptic PDEs

Extended randomized SVD to non-standard Gaussian vectors and Hilbert--Schmidt operators

Designed rational neural networks with higher approximation power for singular functions

Abstract

Discovering hidden partial differential equations (PDEs) and operators from data is an important topic at the frontier between machine learning and numerical analysis. This doctoral thesis introduces theoretical results and deep learning algorithms to learn Green's functions associated with linear partial differential equations and rigorously justify PDE learning techniques. A theoretically rigorous algorithm is derived to obtain a learning rate, which characterizes the amount of training data needed to approximately learn Green's functions associated with elliptic PDEs. The construction connects the fields of PDE learning and numerical linear algebra by extending the randomized singular value decomposition to non-standard Gaussian vectors and Hilbert--Schmidt operators, and exploiting the low-rank hierarchical structure of Green's functions using hierarchical matrices. Rational neural networks (NNs) are introduced and consist of neural networks with trainable rational activation functions. The highly compositional structure of these networks, combined with rational approximation theory, implies that rational functions have higher approximation power than standard activation functions. In addition, rational NNs may have poles and take arbitrarily large values, which is ideal for approximating functions with singularities such as Green's functions. Finally, theoretical results on Green's functions and rational NNs are combined to design a human-understandable deep learning method for discovering Green's functions from data. This approach complements state-of-the-art PDE learning techniques, as a wide range of physics can be captured from the learned Green's functions such as dominant modes, symmetries, and singularity locations.