Furthermore, we performed a Bayesian hyperparameter optimization on the optimizing parameters learning rate and weight decay using the standard Gaussian Process model and Expected Improvement acquisition strategy. As optimization score, we used the equation mentioned above. While the different autoencoders have the same structure, the Bayesian optimization allowed us to use individually optimized hyperparameters for the training process and thus improve the performance. Table \ref{729905} offers an overview of the exact hyperparameters.