The Open UniversitySkip to content
 

The Metric Structure of Weight Space

Rüger, Stefan and Ossen, Arnfried (1997). The Metric Structure of Weight Space. Neural Processing Letters, 5(2) pp. 1–9.

DOI (Digital Object Identifier) Link: http://dx.doi.org/10.1023/A:1009657318698
Google Scholar: Look up in Google Scholar

Abstract

We describe symmetries of feedforward networks in terms of their corresponding groups, which naturally act on and partition weight space. This leads to an algorithm that generates representative weight vectors in a specific fundamental domain. The closure of this domain turns out to be a manifold with singular points. We derive a canonical metric for the manifold that can be implemented efficiently even for large networks. One application would be the clustering of resulting weight vectors of an experiment in order to identify inadequate models or learning methods.

Item Type: Journal Article
ISSN: 1370-4621
Academic Unit/Department: Knowledge Media Institute
Item ID: 11958
Depositing User: Users 8580 not found.
Date Deposited: 08 Oct 2008 13:15
Last Modified: 04 Apr 2011 09:23
URI: http://oro.open.ac.uk/id/eprint/11958
Share this page:

Altmetrics

Scopus Citations

Actions (login may be required)

View Item
Report issue / request change

Policies | Disclaimer

© The Open University   + 44 (0)870 333 4340   general-enquiries@open.ac.uk