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

Download:

Current browse context:

math.DS

Change to browse by:

References & Citations

Bookmark

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

Mathematics > Dynamical Systems

Title: Gradient Descent Only Converges to Minimizers: Non-Isolated Critical Points and Invariant Regions

Abstract: Given a non-convex twice differentiable cost function f, we prove that the set of initial conditions so that gradient descent converges to saddle points where \nabla^2 f has at least one strictly negative eigenvalue has (Lebesgue) measure zero, even for cost functions f with non-isolated critical points, answering an open question in [Lee, Simchowitz, Jordan, Recht, COLT2016]. Moreover, this result extends to forward-invariant convex subspaces, allowing for weak (non-globally Lipschitz) smoothness assumptions. Finally, we produce an upper bound on the allowable step-size.
Comments: 2 figures
Subjects: Dynamical Systems (math.DS); Machine Learning (cs.LG)
Cite as: arXiv:1605.00405 [math.DS]
  (or arXiv:1605.00405v2 [math.DS] for this version)

Submission history

From: Ioannis Panageas [view email]
[v1] Mon, 2 May 2016 09:34:19 GMT (8kb)
[v2] Tue, 7 Jun 2016 07:49:13 GMT (173kb,D)

Link back to: arXiv, form interface, contact.