Real, Esteban, Aggarwal, Alok, Huang, Yanping, Le, Quoc V
The effort devoted to hand-crafting image classifiers has motivated the use
of architecture search to discover them automatically. Although evolutionary
algorithms have been repeatedly applied to architecture search, the
architectures thus discovered have remained inferior to human-crafted ones.
Here we show for the first time that artificially-evolved architectures can
match or surpass human-crafted and RL-designed image classifiers. In
particular, our models---named AmoebaNets---achieved a state-of-the-art
accuracy of 97.87% on CIFAR-10 and top-1 accuracy of 83.1% on ImageNet. Among
mobile-size models, an AmoebaNet with only 5.1M parameters also achieved a
state-of-the-art top-1 accuracy of 75.1% on ImageNet. We also compared this
method against strong baselines. Finally, we performed platform-aware
architecture search with evolution to find a model that trains quickly on
Google Cloud TPUs. This method produced an AmoebaNet that won the Stanford
DAWNBench competition for lowest ImageNet training cost.
Real, Esteban, Aggarwal, Alok, Huang, Yanping, Le, Quoc V
The effort devoted to hand-crafting image classifiers has motivated the use of architecture search to discover them automatically. Although evolutionary algorithms have been repeatedly applied to architecture search, the architectures thus discovered have remained inferior to human-crafted ones. Here we show for the first time that artificially-evolved architectures can match or surpass human-crafted and RL-designed image classifiers. In particular, our models---named AmoebaNets---achieved a state-of-the-art accuracy of 97.87% on CIFAR-10 and top-1 accuracy of 83.1% on ImageNet. Among mobile-size models, an AmoebaNet with only 5.1M parameters also achieved a state-of-the-art top-1 accuracy of 75.1% on ImageNet. We also compared this method against strong baselines. Finally, we performed platform-aware architecture search with evolution to find a model that trains quickly on Google Cloud TPUs. This method produced an AmoebaNet that won the Stanford DAWNBench competition for lowest ImageNet training cost.
https://arxiv.org/abs/1802.01548