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

Download:

Current browse context:

stat.ML

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

Statistics > Machine Learning

Title: Gradient flow dynamics of shallow ReLU networks for square loss and orthogonal inputs

Abstract: The training of neural networks by gradient descent methods is a cornerstone of the deep learning revolution. Yet, despite some recent progress, a complete theory explaining its success is still missing. This article presents, for orthogonal input vectors, a precise description of the gradient flow dynamics of training one-hidden layer ReLU neural networks for the mean squared error at small initialisation. In this setting, despite non-convexity, we show that the gradient flow converges to zero loss and characterise its implicit bias towards minimum variation norm. Furthermore, some interesting phenomena are highlighted: a quantitative description of the initial alignment phenomenon and a proof that the process follows a specific saddle to saddle dynamics.
Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG)
Cite as: arXiv:2206.00939 [stat.ML]
  (or arXiv:2206.00939v2 [stat.ML] for this version)

Submission history

From: Etienne Boursier [view email]
[v1] Thu, 2 Jun 2022 09:01:25 GMT (1871kb,D)
[v2] Mon, 31 Oct 2022 16:42:53 GMT (2754kb,D)

Link back to: arXiv, form interface, contact.