Depth Fragility and Skeletal Universality: Decoupling Topology and Function in Deep Neural Networks

Deep neural networks (DNNs) are traditionally analyzed as black-box function approximators, yet their internal structure exhibits phase transitions characteristic of complex physical systems. In this study, we investigate <i>topological–functional decoupling</i>—the phenomenon whereby a network reta...

Celý popis

Podrobná bibliografie

Hlavní autoři:	Quang Nguyen, Hai Ha Pham, Davide Cassi, Michele Bellingeri
Médium:	Artigo
Jazyk:	Inglês
Vydáno:	MDPI AG 2026-04-01
Edice:	Mathematics
Témata:	deep learning percolation theory network robustness functional skeleton depth fragility
On-line přístup:	https://www.mdpi.com/2227-7390/14/9/1438
Tagy:	Přidat tag Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!