2018luyi / bigan Goto Github PK

System requirements: Python 3.7.1 Tensorflow 2.2.0 PyTorch 1.2.0 GPU - NVIDIA GeForce RTX 2080

JupyterNotebook: 6.0.3 ===============================Data============================================= All data initial files are in data folder. AirQuality- data->air->initial MIMIC- data->mimic->initial

===============================Data Preprocessing=============================== Run data_preporcessAir.ipynb for preprocessing AirQuality data Run data_preporcessMimic.ipynb for preprocessing AirQuality data All data preprocessed files are in data folder. AirQuality- data->air->preprocess MIMIC- data->mimic->preprocess

===============================Bi-GAN============================================= To run Bi-GAN for EHR dataset - Run "biGan/main_ganOrig.ipynb" For training model- Set train=True

For Imputation Testing model-
	Set evalImp=True
	Set missingRate=10 or 20 or 30 or 40 or 50

For Prediction Testing model-
	Set evalPred=True
	Set pred_len=8 or 7 or 6 or 5

To run Bi-GAN for air quality and MIMIC - Run "biGan/main_ganOrigActivity.ipynb" Input Arguments to be set - For AirQuality dataset, set air=True For MIMIC dataset, set mimic=True

For training model-
	Set train=True

For Imputation Testing model-
	Set evalImp=True
	Set missingRate=10 or 20 or 30 or 40 or 50

For Prediction Testing model-
	Set evalPred=True
	Set pred_len=8 or 7 or 6 or 5

===============================BRITS-I============================================= BRITS-I To run BRITS-I for EHR dataset - Run "britsI/main - original.ipynb" For training model- Set train=True

For Imputation Testing model-
	Set evalImp=True
	Set missingRate=10 or 20 or 30 or 40 or 50

For Prediction Testing model-
	Set evalPred=True
	Set pred_len=8 or 7 or 6 or 5

To run BRITS-I for air quality and MIMIC - Run "britsI/main_original_activity.ipynb" Input Arguments to be set - For AirQuality dataset, set air=True For MIMIC dataset, set mimic=True

For training model-
	Set train=True

For Imputation Testing model-
	Set evalImp=True
	Set missingRate=10 or 20 or 30 or 40 or 50

For Prediction Testing model-
	Set evalPred=True
	Set pred_len=8 or 7 or 6 or 5

===============================Baseline============================================= Baseline To run Bi-GAN for EHR dataset - To run Bi-GAN for air quality and MIMIC - MICE - Run "baseline/MICE.ipynb" Input Arguments to be set - For AirQuality dataset, set air=True For MIMIC dataset, set mimic=True For EHR dataset, set ehr=True

For Imputation Testing model-
	Set imp=True
	Set missingRate=10 or 20 or 30 or 40 or 50

For Prediction Testing model-
	Set pred=True
	Set pred_len=8 or 7 or 6 or 5

KNN- Run "baseline/knn.ipynb" Input Arguments to be set - For AirQuality dataset, set air=True For MIMIC dataset, set mimic=True For EHR dataset, set ehr=True

For Imputation Testing model-
	Set imp=True
	Set missingRate=10 or 20 or 30 or 40 or 50

For Prediction Testing model-
	Set pred=True
	Set pred_len=8 or 7 or 6 or 5

MEAN- Run "baseline/mean.ipynb" Input Arguments to be set - For AirQuality dataset, set air=True For MIMIC dataset, set mimic=True For EHR dataset, set ehr=True

For Imputation Testing model-
	Set imp=True
	Set missingRate=10 or 20 or 30 or 40 or 50

For Prediction Testing model-
	Set pred=True
	Set pred_len=8 or 7 or 6 or 5

===============================Bi-GAN Components============================================= Bi-GAN with Wasserstein Loss============================================= To run biWgan for EHR dataset - Run "biWgan/main_Wgan.ipynb.ipynb" For training model- Set train=True

For Imputation Testing model-
	Set evalImp=True
	Set missingRate=10 or 20 or 30 or 40 or 50

For Prediction Testing model-
	Set evalPred=True
	Set pred_len=8 or 7 or 6 or 5

To run biWgan for air quality and MIMIC - Run "biWgan/main_WganActivity.ipynb" Input Arguments to be set - For AirQuality dataset, set air=True For MIMIC dataset, set mimic=True

For training model-
	Set train=True

For Imputation Testing model-
	Set evalImp=True
	Set missingRate=10 or 20 or 30 or 40 or 50

For Prediction Testing model-
	Set evalPred=True
	Set pred_len=8 or 7 or 6 or 5

Bi-GAN without Lambda============================================= To run lambda for EHR dataset - Run "lambda/main_ganLambda.ipynb" For training model- Set train=True

For Imputation Testing model-
	Set evalImp=True
	Set missingRate=10 or 20 or 30 or 40 or 50

For Prediction Testing model-
	Set evalPred=True
	Set pred_len=8 or 7 or 6 or 5

To run lambda for air quality and MIMIC - Run "lambda/main_ganLambdaActivity.ipynb" Input Arguments to be set - For AirQuality dataset, set air=True For MIMIC dataset, set mimic=True

For training model-
	Set train=True

For Imputation Testing model-
	Set evalImp=True
	Set missingRate=10 or 20 or 30 or 40 or 50

For Prediction Testing model-
	Set evalPred=True
	Set pred_len=8 or 7 or 6 or 5

NOTE: Change path of files as required