Mining GOLD Samples for Conditional GANs
PyTorch implementation of "Mining GOLD Samples for Conditional GANs" (NeurIPS 2019).
Run experiments
Run example re-weighting experiments
python main.py --name reweight_base --dataset mnist --epochs 20 --mode acgan_semi
python main.py --name reweight_gold --dataset mnist --epochs 20 --mode acgan_semi_gold
Run rejection sampling experiments
See rejection.ipynb
Run active learning experiments
python main.py --name active_base --dataset mnist --init_size 10 --per_size 2 --max_size 18 --mode acgan_semi --lambda_C_fake 0.01 --query_type random
python main.py --name active_gold --dataset mnist --init_size 10 --per_size 2 --max_size 18 --mode acgan_semi --lambda_C_fake 0.01 --query_type gold
Citation
If you use this code for your research, please cite our papers.
@inproceedings{
mo2019mining,
title={Mining GOLD Samples for Conditional GANs},
author={Mo, Sangwoo and Kim, Chiheon and Kim, Sungwoong and Cho, Minsu and Shin, Jinwoo},
booktitle={Advances in Neural Information Processing Systems},
year={2019},
}