Skip to main content
Cornell University
We gratefully acknowledge support from the Simons Foundation, member institutions, and all contributors. Donate
arxiv logo > stat > arXiv:2006.07814

Help | Advanced Search

arXiv logo
Cornell University Logo

quick links

  • Login
  • Help Pages
  • About

Statistics > Machine Learning

arXiv:2006.07814 (stat)
[Submitted on 14 Jun 2020 (v1), last revised 29 Mar 2021 (this version, v4)]

Title:The Spectrum of Fisher Information of Deep Networks Achieving Dynamical Isometry

Authors:Tomohiro Hayase, Ryo Karakida
View a PDF of the paper titled The Spectrum of Fisher Information of Deep Networks Achieving Dynamical Isometry, by Tomohiro Hayase and 1 other authors
View PDF
Abstract:The Fisher information matrix (FIM) is fundamental to understanding the trainability of deep neural nets (DNN), since it describes the parameter space's local metric. We investigate the spectral distribution of the conditional FIM, which is the FIM given a single sample, by focusing on fully-connected networks achieving dynamical isometry. Then, while dynamical isometry is known to keep specific backpropagated signals independent of the depth, we find that the parameter space's local metric linearly depends on the depth even under the dynamical isometry. More precisely, we reveal that the conditional FIM's spectrum concentrates around the maximum and the value grows linearly as the depth increases. To examine the spectrum, considering random initialization and the wide limit, we construct an algebraic methodology based on the free probability theory. As a byproduct, we provide an analysis of the solvable spectral distribution in two-hidden-layer cases. Lastly, experimental results verify that the appropriate learning rate for the online training of DNNs is in inverse proportional to depth, which is determined by the conditional FIM's spectrum.
Comments: Accepted into AISTATS2021
Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG); Probability (math.PR)
MSC classes: 68T07, 46L54, 60B20, 62E20
ACM classes: G.3
Cite as: arXiv:2006.07814 [stat.ML]
  (or arXiv:2006.07814v4 [stat.ML] for this version)
  https://doi.org/10.48550/arXiv.2006.07814
arXiv-issued DOI via DataCite

Submission history

From: Tomohiro Hayase [view email]
[v1] Sun, 14 Jun 2020 06:32:46 UTC (1,436 KB)
[v2] Tue, 7 Jul 2020 03:29:22 UTC (1,434 KB)
[v3] Mon, 27 Jul 2020 09:36:24 UTC (1,434 KB)
[v4] Mon, 29 Mar 2021 19:08:02 UTC (5,227 KB)
Full-text links:

Access Paper:

    View a PDF of the paper titled The Spectrum of Fisher Information of Deep Networks Achieving Dynamical Isometry, by Tomohiro Hayase and 1 other authors
  • View PDF
  • TeX Source
  • Other Formats
view license
Current browse context:
stat.ML
< prev   |   next >
new | recent | 2020-06
Change to browse by:
cs
cs.LG
math
math.PR
stat

References & Citations

  • NASA ADS
  • Google Scholar
  • Semantic Scholar
a export BibTeX citation Loading...

BibTeX formatted citation

×
Data provided by:

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?)
  • Author
  • Venue
  • Institution
  • Topic

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?)
  • About
  • Help
  • contact arXivClick here to contact arXiv Contact
  • subscribe to arXiv mailingsClick here to subscribe Subscribe
  • Copyright
  • Privacy Policy
  • Web Accessibility Assistance
  • arXiv Operational Status
    Get status notifications via email or slack