Figure 8
From: ClinicalGAN: powering patient monitoring in clinical trials with patient digital twins
Data variation explained by different generative models with TimeGAN (left), CCGAN (−AuxC) (center) and CGAN(+AuxC) (right). We note mode collapse in TimeGAN and CCGAN (−AuxC) but not in CGAN(+AuxC). This is also reflected in the Alpha-Pr scores: TimeGAN (0.02, CGAN(−AuxC) (0.46) and CGAN(+AuxC) (0.77).
