Skip to content

Multi-Objective Neural Architecture Search Based on Diverse Structures and Adaptive Recommendation

Notifications You must be signed in to change notification settings

wangcn0201/MoARR

Repository files navigation

MoARR: Multi-Objective Neural Architecture Search Based on Diverse Structures and Adaptive Recommendation

Multi-Objective Neural Architecture Search Based on Diverse Structures and Adaptive Recommendation
Chunnan Wang, Hongzhi Wang, Guocheng Feng, Fei Geng
https://arxiv.org/abs/2007.02749

Requirements

Python >= 3.5.5, PyTorch >= 1.1.0, torchvision >= 0.2.0, CUDA >= 10.0, cuDNN >= 7.5

Training the CNN code on CIFAR-10

sh Cifar10_DartsTrain.sh

  • Training the given CNN code for 600 epochs using DARTS's setting.

Reproducing the results on CIFAR-10

sh Cifar10_ModelLoad.sh

  • Expected result: 2.48% test error rate with 2.1M model parameters and 0.38G flops.
  • Expected result: 2.62% test error rate with 1.3M model parameters and 0.33G flops.

About

Multi-Objective Neural Architecture Search Based on Diverse Structures and Adaptive Recommendation

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published