GAN Implicit likelihood Estimation

PyTorch demo implementation of paper [Estimate the Implicit Likelihoods of GANs with Application to Anomaly Detection, WWW2020]

Citation

@inproceedings{ren2020estimate,
  title={Estimate the Implicit Likelihoods of GANs with Application to Anomaly Detection},
  author={Ren, Shaogang and Li, Dingcheng and Zhou, Zhixin and Li, Ping},
  booktitle={Proceedings of The Web Conference 2020},
  pages={2287--2297},
  year={2020}
}

Simulation data

step1:

cd model_simulation/

step2:

python3 sim_2_6.py

step3:

python3 main_sim.py --batch_size=50 --z_dim=2 --lrD=0.000004 --lrG=0.000004 --lrIG=0.000004 --input_size=6 --iter_gan=30000 --gpu_ids=0

Or

nohup python3 main_sim.py --batch_size=50 --z_dim=2 --lrD=0.000004 --lrG=0.000004 --lrIG=0.000004 --input_size=6 --iter_gan=30000 --gpu_ids=0 &

Arrhythmia data

step1:

cd model_arrhythmia/

step2:

Downlowd ALAD(https://github.com/houssamzenati/Adversarially-Learned-Anomaly-Detection), unzip and change the folder name to 'ALAD'.

step3:

python3 main_arrhythmia.py --batch_size=30 --z_dim=50 --h_dim=128 --gpu_ids=0  --input_size=274 --iter_gan=1000000 --llk_way=eig --dataset=arrhythmia --lrD=0.000004 --lrG=0.0000004 --lrIG=0.0000004 --out_dir=output

Or

nohup python3 main_arrhythmia.py --batch_size=30 --z_dim=50 --h_dim=128 --gpu_ids=0  --input_size=274 --iter_gan=1000000 --llk_way=eig --dataset=arrhythmia --lrD=0.000004 --lrG=0.0000004 --lrIG=0.0000004 --out_dir=output  &