DNAD: Differentiable Neural Architecture Distillation

To meet the demand for designing efficient neural networks with appropriate trade-offs between model performance (e.g., classification accuracy) and computational complexity, the differentiable neural architecture distillation (DNAD) algorithm is developed based on two cores, namely search by deleti...

Descrizione completa

Salvato in:
Dettagli Bibliografici
Autori principali: Rao, Xuan, Zhao, Bo, Liu, Derong
Natura: Articolo
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2504.20080
http://arxiv.org/abs/2504.20080
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!