Learning Proposals for Probabilistic Programs with Inference Combinators
This provides a flexible framework for researchers and practitioners in probabilistic programming to design tailored variational inference methods, though it appears incremental as it builds on existing concepts like importance sampling and neural networks.
The authors tackled the problem of constructing proposals for probabilistic programs by developing inference combinators, a grammar for composing importance samplers with neural network parameterization, resulting in a correct-by-construction framework for programmable variational methods.
We develop operators for construction of proposals in probabilistic programs, which we refer to as inference combinators. Inference combinators define a grammar over importance samplers that compose primitive operations such as application of a transition kernel and importance resampling. Proposals in these samplers can be parameterized using neural networks, which in turn can be trained by optimizing variational objectives. The result is a framework for user-programmable variational methods that are correct by construction and can be tailored to specific models. We demonstrate the flexibility of this framework by implementing advanced variational methods based on amortized Gibbs sampling and annealing.