We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Improving physics-informed neural networks with meta-learned optimization

Authors: Alex Bihlo
Abstract: We show that the error achievable using physics-informed neural networks for solving systems of differential equations can be substantially reduced when these networks are trained using meta-learned optimization methods rather than to using fixed, hand-crafted optimizers as traditionally done. We choose a learnable optimization method based on a shallow multi-layer perceptron that is meta-trained for specific classes of differential equations. We illustrate meta-trained optimizers for several equations of practical relevance in mathematical physics, including the linear advection equation, Poisson's equation, the Korteweg--de Vries equation and Burgers' equation. We also illustrate that meta-learned optimizers exhibit transfer learning abilities, in that a meta-trained optimizer on one differential equation can also be successfully deployed on another differential equation.
Comments: 15 pages, 10 figures
Subjects: Machine Learning (cs.LG); Numerical Analysis (math.NA)
Cite as: arXiv:2303.07127 [cs.LG]
  (or arXiv:2303.07127v2 [cs.LG] for this version)

Submission history

From: Alexander Bihlo [view email]
[v1] Mon, 13 Mar 2023 13:58:03 GMT (715kb,D)
[v2] Tue, 14 Mar 2023 13:26:03 GMT (920kb,D)

Link back to: arXiv, form interface, contact.