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


Current browse context:


Change to browse by:

References & Citations


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

Mathematics > Numerical Analysis

Title: A multiscale neural network based on hierarchical nested bases

Abstract: In recent years, deep learning has led to impressive results in many fields. In this paper, we introduce a multi-scale artificial neural network for high-dimensional non-linear maps based on the idea of hierarchical nested bases in the fast multipole method and the $\mathcal{H}^2$-matrices. This approach allows us to efficiently approximate discretized nonlinear maps arising from partial differential equations or integral equations. It also naturally extends our recent work based on the generalization of hierarchical matrices [Fan et al. arXiv:1807.01883] but with a reduced number of parameters. In particular, the number of parameters of the neural network grows linearly with the dimension of the parameter space of the discretized PDE. We demonstrate the properties of the architecture by approximating the solution maps of non-linear Schr{\"o}dinger equation, the radiative transfer equation, and the Kohn-Sham map.
Comments: 21 figures, 25 pages. arXiv admin note: text overlap with arXiv:1807.01883
Subjects: Numerical Analysis (math.NA)
Cite as: arXiv:1808.02376 [math.NA]
  (or arXiv:1808.02376v3 [math.NA] for this version)

Submission history

From: Yuwei Fan [view email]
[v1] Sat, 4 Aug 2018 22:54:48 GMT (535kb,D)
[v2] Wed, 26 Sep 2018 22:19:26 GMT (535kb,D)
[v3] Mon, 25 Feb 2019 22:05:52 GMT (632kb,D)

Link back to: arXiv, form interface, contact.