Variational Inference with Normalizing Flows ATML Group 9 - HT22 Repository containing code for the reproducibility challenge set as an exam in the ATML HT22 course. A reproduction of the results of the original paper by Danilo Jimenez Rezende and Shakir Mohamed. The code for the Sylvester flows in flows.py is adapted from https://github.com/riannevdberg/sylvester-flows. Examples of how to train and evaluate the models are found in the Jupyter notebooks train_MNIST and train_CIFAR. Model architecture Effect of the flow-length on MNIST Effect of planar and radial normalising flows on two standard distributions Approximating complex 2D distributions Original input ${x}$ and the reconstructed $\hat{x}$
