Overview

Neural Dawn is an early generative-model project focused on Restricted Boltzmann Machines and Deep Belief Networks.

It combines review notes with PyTorch implementation work: MCMC-based maximum likelihood, Contrastive Divergence, conditional generation, and RBM-based classification experiments.

Why It Fits

The project belongs in the course-project archive as a marker of the older generative-model path: before diffusion and flow matching became the dominant language, energy-based models gave a different way to think about learning distributions.