Neural Architecture Search On Nas Bench 201

Métriques

Accuracy (Test)

Accuracy (Val)

Search time (s)

Résultats

Résultats de performance de divers modèles sur ce benchmark

Nom du modèle	Accuracy (Test)	Accuracy (Val)	Search time (s)	Paper Title	Repository
DiNAS	45.41	46.66	15.36	Multi-conditioned Graph Diffusion for Neural Architecture Search
ParZC	46.34	46.37	-	ParZC: Parametric Zero-Cost Proxies for Efficient NAS	-
AG-Net	46.42	46.73	-	Learning Where To Look -- Generative NAS is Surprisingly Efficient
GDAS	41.71	-	28926	Searching for A Robust Neural Architecture in Four GPU Hours
α-DARTS	46.34	46.17	-	$α$ DARTS Once More: Enhancing Differentiable Architecture Search by Masked Image Modeling	-
EPE-NAS (N=100)	38.80	-	20.5	EPE-NAS: Efficient Performance Estimation Without Training for Neural Architecture Search
NASBOT	46.37	-	75600	Neural Architecture Search with Bayesian Optimisation and Optimal Transport
DARTS-	45.12	44.87	-	DARTS-: Robustly Stepping out of Performance Collapse Without Indicators
arch2vec	46.27	-	-	Does Unsupervised Architecture Representation Learning Help Neural Architecture Search?
Shapley-NAS	46.85	46.57	-	Shapley-NAS: Discovering Operation Contribution for Neural Architecture Search
GEA	46.04	-	-	Guided Evolutionary Neural Architecture Search With Efficient Performance Estimation	-
iDARTS	40.89	40.38	-	iDARTS: Differentiable Architecture Search with Stochastic Implicit Gradients
PNAS +	-	44.75	-	Progressive Neural Architecture Search
DARTS-SaBN	45.85	-	-	Sandwich Batch Normalization: A Drop-In Replacement for Feature Distribution Heterogeneity
DARTS (second order)	16.43	-	29902	DARTS: Differentiable Architecture Search
Λ-DARTS	46.34	46.37	-	$Λ$-DARTS: Mitigating Performance Collapse by Harmonizing Operation Selection among Cells
Local search	46.38	-	151200	Exploring the Loss Landscape in Neural Architecture Search
CATCH-meta	-	46.07	18000	CATCH: Context-based Meta Reinforcement Learning for Transferrable Architecture Search	-
GAEA DARTS (ERM)	46.36	-	-	Geometry-Aware Gradient Algorithms for Neural Architecture Search
NAS without training (N=10)	38.33	-	1.7	Neural Architecture Search without Training

0 of 49 row(s) selected.