Glossary · Phase 4 · Deep learning
Reconstruction error
How far an autoencoder's output is from its input; high for items unlike the data it trained on, which makes it an anomaly score.
Try it now
Each use leans on a different part: the code, the reconstruction, or the error.
Learn it properly
Autoencoders · Encoder, bottleneck and decoder, reconstruction error for anomalies, denoising, and the latent space image generators use. About 35 minutes.
Other terms from the same lesson
- autoencoder: A network trained to reconstruct its own input through a narrow middle layer, which learns a compressed code of the data.
- latent space: The space of an autoencoder's codes; image generators like Stable Diffusion work there instead of on pixels.
Never trained an AI?
Teach a small one to write, right in your browser. It takes about 20 seconds and needs no account.