Autoencoder

An autoencoder is a neural network trained to reproduce its input after compressing it into a smaller internal representation. The encoder converts the original data into a latent code, and the decoder reconstructs the input from that code. Because the model must preserve useful structure while discarding less important detail, autoencoders support dimensionality reduction, denoising, feature learning, compression, and anomaly detection. A reconstruction that differs greatly from the input may indicate an unfamiliar pattern. Variants can generate new samples, but the compressed representation is shaped by the training data and objective, so it may also preserve bias or omit information needed for another task.