Autoencoder

Created: 2022-03-24 10:16
#note

An autoencoder is a neural network that reconstructs its input data in the output layer-> the encoder maps the input to an internal representation and the decoder that maps this representation to a reconstruction of the input. This architecture contrasts with Adversarial Networks, which use competing network objectives.

Pros: #dimensionality_reduction, #data_reconstruction, #feature_extraction

Two approaches:

  • usning autoencoders to learn lower-dimensional feature representations at the bottleneck layer;
  • filling the blanks of the interaction matrix directly in the reconstruction layer

Types:

  • Denoising AE;
  • Variational AE;
  • Contractive AE;
  • Marginalized AE

References

  1. https://medium.com/sciforce/deep-learning-based-recommender-systems-b61a5ddd5456

Tags

#autoencoder