Important About Undercomplete Autoencoder
Autoencoders in a broader sense: If we were to sum them up in a single statement, it would probably go like this: Yes, it may appear to be simple and pointless. We will see, however, that this is not trivial nor insignificant. Autoencoders , in reality, are deep models that can learn dense representations of the input. Latent representations or coding are the names given to these representations. There are two pieces to an Automatic encoder: An encoder: It is a component of the model that accepts input data and compresses it. Where x is the input data, c is the latent representation, and E is our encoding function, E(x) = c A decoder: This component accepts the latent representation as a parameter and attempts to recreate the original input. D(c) = x’, where x’ is the decoder’s output and D is our decoding function Undercomplete Autoencoder Our network should be able to learn how to reassemble our input data throughout the training phase. The Autoencoders model architecture is show...