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

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

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

Computer Science > Machine Learning

Title: Robust Training and Initialization of Deep Neural Networks: An Adaptive Basis Viewpoint

Abstract: Motivated by the gap between theoretical optimal approximation rates of deep neural networks (DNNs) and the accuracy realized in practice, we seek to improve the training of DNNs. The adoption of an adaptive basis viewpoint of DNNs leads to novel initializations and a hybrid least squares/gradient descent optimizer. We provide analysis of these techniques and illustrate via numerical examples dramatic increases in accuracy and convergence rate for benchmarks characterizing scientific applications where DNNs are currently used, including regression problems and physics-informed neural networks for the solution of partial differential equations.
Comments: 26 pages
Subjects: Machine Learning (cs.LG); Numerical Analysis (math.NA); Machine Learning (stat.ML)
Cite as: arXiv:1912.04862 [cs.LG]
  (or arXiv:1912.04862v1 [cs.LG] for this version)

Submission history

From: Mamikon Gulian [view email]
[v1] Tue, 10 Dec 2019 18:04:03 GMT (1915kb,D)

Link back to: arXiv, form interface, contact.