Skip to main content
We gratefully acknowledge support from the Simons Foundation, member institutions, and all contributors. Donate
arXiv logo
Cornell University Logo

Statistics > Machine Learning

arXiv:1803.00885 (stat)
[Submitted on 2 Mar 2018 (v1), last revised 22 Feb 2019 (this version, v5)]

Title:Essentially No Barriers in Neural Network Energy Landscape

Authors:Felix Draxler, Kambis Veschgini, Manfred Salmhofer, Fred A. Hamprecht
View PDF
Abstract:Training neural networks involves finding minima of a high-dimensional non-convex loss function. Knowledge of the structure of this energy landscape is sparse. Relaxing from linear interpolations, we construct continuous paths between minima of recent neural network architectures on CIFAR10 and CIFAR100. Surprisingly, the paths are essentially flat in both the training and test landscapes. This implies that neural networks have enough capacity for structural changes, or that these changes are small between minima. Also, each minimum has at least one vanishing Hessian eigenvalue in addition to those resulting from trivial invariance.
Comments: In Proceedings of 35th International Conference on Machine Learning (ICML 2018)
Subjects: Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
Cite as: arXiv:1803.00885 [stat.ML]
  (or arXiv:1803.00885v5 [stat.ML] for this version)
  https://doi.org/10.48550/arXiv.1803.00885
Journal reference: Proceedings of the 35th International Conference on Machine Learning, PMLR 80:1308-1317, 2018

Submission history

From: Felix Draxler [view email]
[v1] Fri, 2 Mar 2018 15:22:10 UTC (760 KB)
[v2] Thu, 8 Mar 2018 21:59:03 UTC (1,548 KB)
[v3] Thu, 22 Mar 2018 17:45:05 UTC (250 KB)
[v4] Mon, 25 Jun 2018 14:55:25 UTC (3,689 KB)
[v5] Fri, 22 Feb 2019 11:20:22 UTC (3,625 KB)
Full-text links:

Access Paper:

  • View PDF
  • TeX Source
  • Other Formats
license icon view license
Current browse context:
cs
< prev   |   next >
new | recent | 2018-03
Change to browse by:
cs.AI
cs.LG
stat
stat.ML

References & Citations

export BibTeX citation

Bookmark

BibSonomy logo Reddit logo

Bibliographic and Citation Tools

Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)

Code, Data and Media Associated with this Article

alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)

Demos

Replicate (What is Replicate?)
Hugging Face Spaces (What is Spaces?)
TXYZ.AI (What is TXYZ.AI?)

Recommenders and Search Tools

Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)