ZhuSuan: A Library for Bayesian Deep Learning

TL;DR

ZhuSuan is a Python library built on TensorFlow that enables Bayesian deep learning by supporting various probabilistic models and inference methods, bridging the gap between Bayesian inference and deep neural networks.

Contribution

It introduces ZhuSuan, a novel probabilistic programming library that integrates Bayesian methods with deep learning, supporting diverse models and inference techniques.

Findings

Supports a wide range of probabilistic models including deep generative models.

Facilitates Bayesian inference in deep neural networks using TensorFlow.

Provides practical examples like Bayesian logistic regression and variational auto-encoders.

Abstract

In this paper we introduce ZhuSuan, a python probabilistic programming library for Bayesian deep learning, which conjoins the complimentary advantages of Bayesian methods and deep learning. ZhuSuan is built upon Tensorflow. Unlike existing deep learning libraries, which are mainly designed for deterministic neural networks and supervised tasks, ZhuSuan is featured for its deep root into Bayesian inference, thus supporting various kinds of probabilistic models, including both the traditional hierarchical Bayesian models and recent deep generative models. We use running examples to illustrate the probabilistic programming on ZhuSuan, including Bayesian logistic regression, variational auto-encoders, deep sigmoid belief networks and Bayesian recurrent neural networks.