Name
Abstract | ||
|---|---|---|
We describe symmetries of feedforward networks in terms of theircorresponding groups, which naturally act on and partition weight space.This leads to an algorithm that generates representative weight vectors ina specific fundamental domain. The closure of this domain turns out to bea manifold with singular points. We derive a canonical metric for themanifold that can be implemented efficiently even for large networks. Oneapplication would be the clustering of resulting weight vectors of anexperiment in order to identify inadequate models or learning methods. |
Year | DOI | Venue |
|---|---|---|
1997 | 10.1023/A:1009657318698 | Neural Processing Letters |
Keywords | Field | DocType |
canonical metric,clustering in weight space,fundamental domain,symmetries,weight space | Topology,Fisher information metric,Pseudometric space,Convex metric space,Intrinsic metric,Metric (mathematics),Statistical manifold,Pseudo-Riemannian manifold,Mathematics,Injective metric space | Journal |
Volume | Issue | ISSN |
5 | 2 | 1573-773X |
Citations | PageRank | References |
2 | 0.42 | 4 |
Authors | ||
2 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Stefan M. Rüger | 1 | 499 | 51.53 |
| Arnfried Ossen | 2 | 7 | 1.72 |