This Git-repository includes source code to run experiments and generate figures for
Reparameterization Gradients through Acceptance-Rejection Sampling Algorithms. Christian A. Naesseth, Francisco J. R. Ruiz, Scott W. Linderman, and David M. Blei Proceedings of the 20th International Conference on Artificial Intelligence and Statistics 2017, Fort Lauderdale, Florida, USA.
Jupyter notebook to generate Figure 1 in the paper.
Jupyter notebooks to generate Figure 2 and 3.
Jupyter notebooks and Python code to run experiments for the Dirichlet-Multionomial example, as well as generating Figure 4.
sparse gamma def
Jupyter notebooks and Python code to run experiments for the Sparse Gamma DEF example. Also includes results and data used to generate Figure 5 and Table 1.