This vignette illustrates the intended role of autoencoders within BioMoR.
Currently, the autoencoder functionality is implemented as a lightweight stub that returns placeholder embeddings. This allows users to experiment with the high-level pipeline without requiring heavy deep-learning dependencies.