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

Download:

Current browse context:

math.NA

Change to browse by:

References & Citations

Bookmark

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

Mathematics > Numerical Analysis

Title: A randomized Newton's method for solving differential equations based on the neural network discretization

Abstract: We develop a randomized Newton's method for solving differential equations, based on a fully connected neural network discretization. In particular, the randomized Newton's method randomly chooses equations from the overdetermined nonlinear system resulting from the neural network discretization and solves the nonlinear system adaptively. We prove theoretically that the randomized Newton's method has a quadratic convergence locally. We also apply this new method to various numerical examples, from one- to high-dimensional differential equations, in order to verify its feasibility and efficiency. Moreover, the randomized Newton's method can allow the neural network to "learn" multiple solutions for nonlinear systems of differential equations, such as pattern formation problems, and provides an alternative way to study the solution structure of nonlinear differential equations overall.
Subjects: Numerical Analysis (math.NA); Optimization and Control (math.OC)
Cite as: arXiv:1912.03196 [math.NA]
  (or arXiv:1912.03196v1 [math.NA] for this version)

Submission history

From: Qipin Chen [view email]
[v1] Fri, 6 Dec 2019 16:01:25 GMT (1596kb,D)

Link back to: arXiv, form interface, contact.