MachineLearning/autoencoder library

🧠 Autoencoder (simple dense feed-forward encoder/decoder)

A small, configurable autoencoder using fully-connected layers and mean squared error reconstruction loss. Designed to be explicit and easy to understand with clear docs for engineering adaptation.

Contract:

  • Input: X (n x m), hidden dimension h.
  • Output: encoded representations and reconstruction function.
  • Error: throws for invalid dimensions.

Time Complexity: O(n * m * h * iters) Space Complexity: O(m * h)

Classes

Autoencoder