Searching for A Robust Neural Architecture in Four GPU Hours

We propose A Gradient-based neural architecture search approach using Differentiable Architecture Sampler (GDAS).

Figure-1. We utilize a DAG to represent the search space of a neural cell. Different operations (colored arrows) transform one node (square) to its intermediate features (little circles). Meanwhile, each node is the sum of the intermediate features transformed from the previous nodes. As indicated by the solid connections, the neural cell in the proposed GDAS is a sampled sub-graph of this DAG. Specifically, among the intermediate features between every two nodes, GDAS samples one feature in a differentiable way.

Requirements

• PyTorch 1.0.1
• Python 3.6
• opencv

conda install pytorch torchvision cuda100 -c pytorch

Usages

Train the searched CNN on CIFAR

CUDA_VISIBLE_DEVICES=0 bash ./scripts-cnn/train-cifar.sh GDAS_FG cifar10  cut
CUDA_VISIBLE_DEVICES=0 bash ./scripts-cnn/train-cifar.sh GDAS_F1 cifar10  cut
CUDA_VISIBLE_DEVICES=0 bash ./scripts-cnn/train-cifar.sh GDAS_V1 cifar100 cut

Train the searched CNN on ImageNet

CUDA_VISIBLE_DEVICES=0 bash ./scripts-cnn/train-imagenet.sh GDAS_F1 52 14
CUDA_VISIBLE_DEVICES=0 bash ./scripts-cnn/train-imagenet.sh GDAS_V1 50 14

Evaluate a trained CNN model

CUDA_VISIBLE_DEVICES=0 python ./exps-cnn/evaluate.py --data_path  $TORCH_HOME/cifar.python --checkpoint ${checkpoint-path}
CUDA_VISIBLE_DEVICES=0 python ./exps-cnn/evaluate.py --data_path  $TORCH_HOME/ILSVRC2012 --checkpoint ${checkpoint-path}

Train the searched RNN

CUDA_VISIBLE_DEVICES=0 bash ./scripts-rnn/train-PTB.sh DARTS_V1
CUDA_VISIBLE_DEVICES=0 bash ./scripts-rnn/train-PTB.sh DARTS_V2
CUDA_VISIBLE_DEVICES=0 bash ./scripts-rnn/train-PTB.sh GDAS
CUDA_VISIBLE_DEVICES=0 bash ./scripts-rnn/train-WT2.sh DARTS_V1
CUDA_VISIBLE_DEVICES=0 bash ./scripts-rnn/train-WT2.sh DARTS_V2
CUDA_VISIBLE_DEVICES=0 bash ./scripts-rnn/train-WT2.sh GDAS

Training Logs

Some training logs can be found in ./data/logs/, and some pre-trained models can be found in Google Driver.

Experimental Results

Figure 2. Top-1 and top-5 errors on ImageNet.

Citation

@inproceedings{dong2019search,
  title={Searching for A Robust Neural Architecture in Four GPU Hours},
  author={Dong, Xuanyi and Yang, Yi},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2019}
}