Edward: A library for probabilistic modeling, inference, and criticism

TL;DR

Edward is a flexible library built on TensorFlow that facilitates probabilistic modeling, inference, and model criticism, supporting scalable and efficient analysis of complex models across hardware like GPUs.

Contribution

The paper introduces Edward, a library that integrates probabilistic modeling, inference, and criticism within TensorFlow, enabling scalable and efficient development of complex models.

Findings

Supports a broad class of probabilistic models

Enables distributed training on GPUs

Facilitates model criticism techniques

Abstract

Probabilistic modeling is a powerful approach for analyzing empirical information. We describe Edward, a library for probabilistic modeling. Edward's design reflects an iterative process pioneered by George Box: build a model of a phenomenon, make inferences about the model given data, and criticize the model's fit to the data. Edward supports a broad class of probabilistic models, efficient algorithms for inference, and many techniques for model criticism. The library builds on top of TensorFlow to support distributed training and hardware such as GPUs. Edward enables the development of complex probabilistic models and their algorithms at a massive scale.