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

Statistics > Machine Learning

Title: Fixed points of arbitrarily deep 1-dimensional neural networks

Abstract: In this paper, we introduce a new class of functions on $\mathbb{R}$ that is closed under composition, and contains the logistic sigmoid function. We use this class to show that any 1-dimensional neural network of arbitrary depth with logistic sigmoid activation functions has at most three fixed points. While such neural networks are far from real world applications, we are able to completely understand their fixed points, providing a foundation to the much needed connection between application and theory of deep neural networks.
Comments: 9 pages, 3 figures
Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG); Dynamical Systems (math.DS)
MSC classes: 37E99 (Primary) 37N99 (Secondary)
Cite as: arXiv:2303.12814 [stat.ML]
  (or arXiv:2303.12814v1 [stat.ML] for this version)

Submission history

From: Andrew Cook [view email]
[v1] Wed, 22 Mar 2023 04:25:28 GMT (82kb,D)

Link back to: arXiv, form interface, contact.