Denoising Autoencoder

Fact

An Autoencoder that learns by fixing a noisy version of its input.

In Plain Words

A Denoising Autoencoder is like cleaning a ketchup-smudged school photo. First we add the ketchup. Then it learns to find the face anyway.

You meet it in pretraining, cleanup, and weird-data spotting. It learns features that still work when the input gets messy.

Related Concepts

• Autoencoder
• Representation Learning
• Self-supervised Learning
• Regularization

Autoencoder
It adds noise to an Autoencoder input, then learns to rebuild it.

Representation Learning
It learns stable features instead of copying the input.

SSL
It makes fake noise, then uses the clean input as the answer.

Regularization
Noise training works like Regularization and reduces memorizing.