AlgoPlus//machine learning / autoencoder
Read the theory

Autoencoder

Compress the input through a narrow bottleneck, then reconstruct it.

Phase
Recon error
Legend
Input
Latent code
Reconstruction
AI Tutor Workspace
In a nutshell
An autoencoder is a network trained to copy its input to its output through a narrow middle layer. The encoder compresses the input into a small latent code (the bottleneck); the decoder rebuilds the input from that code. Because the code is far smaller than the input, the network can't just memorize — it must learn the data's essential structure, which makes autoencoders useful for compression, denoising, and unsupervised feature learning.
Ready
Press play to begin the cinematic walkthrough.
Force the data through a keyhole and back. To rebuild what it sees, the network has to throw away the noise and keep only the gist — compression as a side effect of copying.
Key terms
Go deeper in the lesson
Read the full theory, intuition & complexity for Autoencoder.