The practical value is illustrated in two ways.

Neural Information Processing Systems http://nips.cc/

We would like to thank the reviewers for their thoughtful comments and valuable suggestions. We will clarify this point in the paper. Our algorithms are agnostic to the leaf distributions used. Thanks for this valuable feedback, we will improve the pseudocode as you suggest. As such, there is memory overhead but no computational overhead.

First, there are often pathologies in LMMs that hinder efficient HMC sampling.

Theorem 2) is optimized (4) to correct the bulk approximation.

Here, we present an online version of Algo. 1, where the term'online' refers to the fact that the

We introduce Support Decomposition V ariational Inference (SDVI), a new varia-tional inference (VI) approach for probabilistic programs with stochastic support.

Figure 7: On the left, a standard probabilistic program is reformatted such that observe statements are replaced with annotations. On the right are 3 executions with randomly masked variables.

In this paper, we propose a complementary approach to scale inference for probabilistic programs.