Deep networks on toroids: removing symmetries reveals the structure of flat regions in the landscape geometry

We systematize the approach to the investigation of deep neural network landscapes by basing it on the geometry of the space of implemented functions rather than the space of parameters. Grouping classifiers into equivalence classes, we develop a standardized parameterization in which all symmetries...

Ausführliche Beschreibung

Bibliographische Detailangaben

Veröffentlicht in:	Journal of statistical mechanics 2022-11, Vol.2022 (11), p.114007
Hauptverfasser:	Pittorino, Fabrizio, Ferraro, Antonio, Perugini, Gabriele, Feinauer, Christoph, Baldassi, Carlo, Zecchina, Riccardo
Format:	Artikel
Sprache:	eng
Online-Zugang:	Volltext
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!