Autoencoder Explained - Deep Neural Networks

#datascience #machinelearning #neuralnetworks

An autoencoder is a neural network that learns to copy its input to its output

An Autoencoder can be divided into two parts: the encoder and the decoder. The encoder is a mapping from the input space into a lower dimensional latent space (bottle neck layer).

At this stage it is nothing but low dimensional representations of data in an unsupervised manner

And what is does here is nothing but dimenstionality reduction similar to what PCA does

the potential of Autoencoders resides in their non-linearity, allowing the model to learn more powerful generalizations compared to PCA, and to reconstruct back the input with a significantly lower loss of information

The decoder is a mapping from the low dimension latent space into the reconstruction space with a dimensionality equal to the input space

The output in reconstruction space is close to similar to input but there is some loss of information this is called as reconstruction error

One potential use case of autoencoders is anomaly detection

This is more useful when we have very few negative cases and classes are imbalnced but it can be used in normal scenrio as well where labelling is hard