Skip to content
Technical GlossaryDeep Learning

Contractive Autoencoder

A type of autoencoder that uses an additional penalty to learn more stable latent representations under input perturbations.

A contractive autoencoder focuses not only on reconstruction quality but also on the stability of latent representations under small input perturbations. Through a Jacobian-based penalty, it encourages learned representations to lie on smoother and more robust manifolds. This can help extract more meaningful latent structure in noisy or sensitive datasets.