Autoencoders

Networks trained to reconstruct their input through a bottleneck, learning compressed representations. Foundational for dimensionality reduction and generative models.