Fig. 4.

Mean and standard deviation of the reconstruction loss during the training of our VAE, plotted for both training and heldout data across 10 different folds. This illustrates the robustness of our approach by showing consistent performance across training data and heldout data