Edward: a Library for Probabilistic Machine Learning and Statistics

• Activity Number: 167 - Statistical Computing and Statistical Graphics: Student Paper Award and Chambers Statistical Software Award
• Type: Contributed
• Date/Time: Monday, July 30, 2018 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Computing
• Author(s): Dustin Tran* and David Blei
• Companies: Columbia University and Columbia University
• Keywords: probabilistic programming; deep learning; probabilistic machine learning; variational inference; neural networks

Abstract

Probabilistic machine learning has expanded the scope of statistical analysis, with applications ranging from perceptual tasks such as image generation, to scientific challenges such as understanding how populations in ecology evolve over time and how genetic factors cause diseases. In this talk, I will provide an overview of Edward, a library for probabilistic machine learning. Edward supports compositions of both models and inference for flexible experimentation, ranging from a variety of composable modeling blocks such as neural networks, graphical models, and probabilistic programs; and a variety of composable inferences such as point estimation, variational inference, and MCMC. As part of TensorFlow, Edward scales training with accelerator support such as GPUs.