Architecture: This model is designed for concurrent training on information across diverse modalities.

Generative core model: At the core this model is a veriational autoencoder that generates both MNIST and Fashion-MNIST data.