Fig. 3

Mean reconstruction error over various latent dimensions indicates as the latent dimension increases, the mean reconstruction error decreases. This shows a trade-off between compression and model accuracy. a Fs-Peptide exhibits a linear decrease in reconstruction error as the latent dimension increases. b VHP begins to show the limits of increasing the model’s latent dimension; the lowest recorded mean reconstruction loss was found to be at latent dimension nine. c While the lowest reconstruction error for BBA was found to be with a latent dimension of size ten, there is a wide amount of variation in the reconstruction loss for this latent dimension size. Various values of the reconstruction error at each latent dimension shows there to be interactions between the various model hyperparameters