Deep Probabilistic Programming

TL;DR

Edward is a flexible, efficient probabilistic programming language integrated with TensorFlow, enabling diverse inference methods and outperforming existing systems in speed for various models.

Contribution

Introduction of Edward, a Turing-complete probabilistic programming language that treats inference as a first-class citizen, enabling flexible modeling and inference with significant computational speedups.

Findings

Edward is at least 35x faster than Stan on logistic regression.

Edward is 6x faster than PyMC3.

No runtime overhead, matching TensorFlow speed.

Abstract

We propose Edward, a Turing-complete probabilistic programming language. Edward defines two compositional representations---random variables and inference. By treating inference as a first class citizen, on a par with modeling, we show that probabilistic programming can be as flexible and computationally efficient as traditional deep learning. For flexibility, Edward makes it easy to fit the same model using a variety of composable inference methods, ranging from point estimation to variational inference to MCMC. In addition, Edward can reuse the modeling representation as part of inference, facilitating the design of rich variational models and generative adversarial networks. For efficiency, Edward is integrated into TensorFlow, providing significant speedups over existing probabilistic systems. For example, we show on a benchmark logistic regression task that Edward is at least 35x faster than Stan and 6x faster than PyMC3. Further, Edward incurs no runtime overhead: it is as fast as handwritten TensorFlow.