PyTorch implementations of FVSBN, NADE, and MADE — foundational autoregressive density estimators for binarised MNIST — with a vectorised NADE training loop, a 15-run hyper-parameter sweep, and publication-quality architecture figures.
-
Updated
Apr 19, 2026 - Python