Rethinking Variational Inference for Probabilistic Programs with Stochastic Support

We introduce Support Decomposition Variational Inference (SDVI), a new variational inference (VI) approach for probabilistic programs with stochastic support. Existing approaches to this problem rely on designing a single global variational guide on a variable-by-variable basis, while maintaining the stochastic control flow of the original program. SDVI instead breaks the program down into sub-programs with static support, before automatically building separate sub-guides for each. This decomposition significantly aids in the construction of suitable variational families, enabling, in turn, substantial improvements in inference performance.

Beyond Bayesian Model Averaging over Paths in Probabilistic Programs with Stochastic Support

The posterior in probabilistic programs with stochastic support decomposes as a weighted sum of the local posterior distributions associated with each possible program path. We show that making predictions with this full posterior implicitly performs a Bayesian model averaging (BMA) over paths. This is potentially problematic, as model misspecification can cause the BMA weights to prematurely collapse onto a single path, leading to sub-optimal predictions in turn. To remedy this issue, we propose alternative mechanisms for path weighting: one based on stacking and one based on ideas from PAC-Bayes. We show how both can be implemented as a cheap post-processing step on top of existing inference engines. In our experiments, we find them to be more robust and lead to better predictions compared to the default BMA weights.

Expectation Programming

Building on ideas from probabilistic programming, we introduce the concept of an expectation programming framework (EPF) that automates the calculation of expectations. Analogous to a probabilistic program, an expectation program is comprised of a mix of probabilistic constructs and deterministic calculations that define a conditional distribution over its variables. However, the focus of the inference engine in an EPF is to directly estimate the resulting expectation of the program return values, rather than approximate the conditional distribution itself. This distinction allows us to achieve substantial performance improvements over the standard probabilistic programming pipeline by tailoring the inference to the precise expectation we care about. We realize a particular instantiation of our EPF concept by extending the probabilistic programming language Turing to allow so-called target-aware inference to be run automatically, and show that this leads to significant empirical gains compared to conventional posterior-based inference.