AutoBNN

This is the implementation of Searching for Accurate Binary Neural Architectures

Network

The implementation of VGG_Small and ResNet18 is in vgg_small.py and resnet18.py

Network Expansion Ratio
VGG-Auto-A [0.5, 2, 1, 1, 1, 0.5]
VGG-Auto-B [2, 2, 4, 2, 4, 0.5]
Res18-Auto-A [2, 4, 2, 2, 2, 4, 4, 4, 4, 4, 4, 4]
Res18-Auto-B [3, 4, 4, 3, 1, 4, 4, 4, 3, 3, 3, 3]

Citation

@inproceedings{shen2019searching, 
    title={Searching for accurate binary neural architectures},
    author={Shen, Mingzhu and Han, Kai and Xu, Chunjing and Wang, Yunhe},
    booktitle={ICCV Neural Architecture Workshop},
    pages={0--0},
    year={2019}
}