variational-autoencoder
The project presents implementations of variational autoencoders using TensorFlow and PyTorch aimed at modeling binarized MNIST images. By incorporating techniques such as inverse autoregressive flow, the models significantly enhance the marginal likelihood estimates. This guide provides comprehensive setup instructions for both PyTorch and Jax, facilitates performance comparison, and outlines procedures for creating visual outputs, highlighting increased inference efficiency and adaptability across computational contexts.
