Architecture
Autoencoder
Network trained to reconstruct its input through a low-dimensional bottleneck.
Definition
An autoencoder compresses an input into a latent code and reconstructs it, learning a compact representation. Variants include denoising and masked autoencoders. They power feature learning, anomaly detection, and the VAE family used in latent-diffusion image models.
Common use cases
- Representation learning
- Anomaly detection
- Image compression