Edward

Edward is an open-source Python library designed for probabilistic modeling, inference, and model criticism, primarily aimed at researchers and practitioners in machine learning, statistics, and probabilistic programming. Built on TensorFlow, it supports a variety of probabilistic modeling paradigms, advanced inference methods (including variational inference and Monte Carlo techniques), compatibility with deep learning libraries such as TensorFlow Layers and Keras, and robust tools for model evaluation such as predictive checks and point-based evaluations. Edward serves as a flexible testbed for fast experimentation with Bayesian, neural, and generative models.