What is an Auto-Encoder?

What is an Auto-Encoder?

Autoencoders are structured to take an input, transform this input into a different representation, an embedding of the input. From this embedding, it aims to reconstruct the original input as precisely as possible. It basically tries to copy the input.

Autoencoders are obtained from unsupervised deep learning algorithm.

We train a deep neural network with a bottleneck, where we keep the input and output identical.

Autoencoders are used for the lower dimensional representation of input features.

Autoencoders are used in following cases -

  • To reduce the dimension of linear and non-linear data. That is the primary reason of its superiority over PCA.
  • It is also used in recommendation systems, they are good at understanding user preferences.
  • They are used to extract important features.
  • Multiple features of an image can be understood by using multiple autoencoders.