There are various types of standard autoencoder. Here we will explain the most widely used ones and go through some coded examples in Keras.
Standard types of autoencoder
Undercomplete autoencoders
Undercomplete autoencoder architectures can be used to constrain the number of nodes that are present in the hidden layers of the network, limiting the amount of information that can flow through it. The model can learn the most important attributes of the input data by penalizing it as per the reconstruction error. This reconstruction error is essentially the difference between the input and the reconstructed output from the encoding. The encoding learns and describes the latent attributes of the input data.