INFO

A type of neural network designed to learn efficient representations of data by compressing input into a lower-dimensional form (encoding) and then reconstructing it back (decoding) as closely as possible to the original.

  • Represent one of the foundational unsupervised deep learning architectures

Components:

  • Compression (Encoder): Transforms input data into a compact latent representation
  • Latent Space (Bottleneck): The compressed form that captures the most essential features
  • Decompression (Decoder): Reconstructs the original data from the latent representation

How it Works

Goal

  • Dimensionality Reduction: distill complex, high-dimensional data into a reduced, meaningful representation or latent space

Reconstruction error

  • Discrepancy between original data and decoded reconstruction
  • When minimized will
    • Uncover intrinsic features
    • Enable businesses to process large datasets efficiently
  • Variants enhances capability
    • Denoising autoencoders: facilitating noise removal
    • Variational autoencoders: probabilistic generation of new data