Lets look at some of its Hyperparameters Autoencoder Specific Compression factor (Latent space size) Loss function (mse?) Neural Network architecture Number of layers Number of neurons in each layer (Shape of the matrices $A_n$) Optimisation parameters Learning Rate Controls how fast the parameters are found To high value makes the training unstable Batch size Controls how many samples are averaged together. Lower values make the training more stable, but also the result less optimal