ProxylessNAS: Direct Neural Architecture Search on Target Task and Hardware

ICLR 2019 Han CaiLigeng ZhuSong Han

Neural architecture search (NAS) has a great impact by automatically designing effective neural network architectures. However, the prohibitive computational demand of conventional NAS algorithms (e.g. $10^4$ GPU hours) makes it difficult to \emph{directly} search the architectures on large-scale tasks (e.g. ImageNet)... (read more)

PDF Abstract
TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK COMPARE
Image Classification CIFAR-10 Proxyless-G + c/o Percentage correct 97.92 # 7
Image Classification CIFAR-10 Proxyless-G + c/o Percentage error 2.08 # 3
Neural Architecture Search CIFAR-10 Image Classification Proxyless-G + c/o Percentage error 2.08 # 1
Neural Architecture Search CIFAR-10 Image Classification Proxyless-G + c/o Params 5.7M # 2