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: Deep neural networks can stably solve high-dimensional, noisy, non-linear inverse problems

Abstract: We study the problem of reconstructing solutions of inverse problems when only noisy measurements are available. We assume that the problem can be modeled with an infinite-dimensional forward operator that is not continuously invertible. Then, we restrict this forward operator to finite-dimensional spaces so that the inverse is Lipschitz continuous. For the inverse operator, we demonstrate that there exists a neural network which is a robust-to-noise approximation of the operator. In addition, we show that these neural networks can be learned from appropriately perturbed training data. We demonstrate the admissibility of this approach to a wide range of inverse problems of practical interest. Numerical examples are given that support the theoretical findings.
Subjects: Numerical Analysis (math.NA); Analysis of PDEs (math.AP); Machine Learning (stat.ML)
MSC classes: 35R30, 41A25, 68T05
Cite as: arXiv:2206.00934 [math.NA]
  (or arXiv:2206.00934v4 [math.NA] for this version)

Submission history

From: Philipp Petersen [view email]
[v1] Thu, 2 Jun 2022 08:51:46 GMT (205kb,D)
[v2] Tue, 7 Jun 2022 15:45:13 GMT (211kb,D)
[v3] Mon, 13 Jun 2022 09:50:25 GMT (211kb,D)
[v4] Tue, 21 Jun 2022 07:57:09 GMT (212kb,D)

Link back to: arXiv, form interface, contact.