This type of autoencoder is designed to contract the input space in the hidden layer, thus making the model more robust to slight changes in input space. 27.07.2023 17:54 aior