Synthetic Data Generation on eICU (test)
0.012MMDAugmented TimeGAN
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Augmented TimeGANepochs=7000, layers=3, latent_dim=24, noise_sd=0.22025.11 | 0.012 | 0.053 | 95.1 | 94.1 | 41.5 | |
| TransFusionepochs=7000, layers=3, latent_dim=24, noise_sd=0.22025.11 | 0.014 | 0.032 | 94.2 | 96.4 | 57.4 | |
| TimeDiffepochs=7000, layers=3, latent_dim=24, noise_sd=0.22025.11 | 0.018 | 0.071 | 69.8 | 71.7 | 53.2 | |
| DP-TimeGANepochs=7000, layers=3, latent_dim=24, noise_sd=0.2, epsilon=20, delta=10^-52025.11 | 0.019 | 0.145 | 89.4 | 92 | 58.1 | |
| DP Normalizing Flowsepochs=7000, layers=3, latent_dim=24, noise_sd=0.2, epsilon=20, delta=10^-52025.11 | 0.02 | 0.167 | 77.6 | 63.7 | 68.4 | |
| SeriesGANepochs=7000, layers=3, latent_dim=24, noise_sd=0.22025.11 | 0.102 | 0.24 | 86.6 | 78.8 | 46.7 |