Deep Probabilistic Programming
Deep Probabilistic Programming
复制标题
DOI：
--
复制
发表时间：
2017-01
期刊：
影响因子：
--
通讯作者：
Dustin Tran;M. Hoffman;R. Saurous;E. Brevdo;K. Murphy;D. Blei
中科院分区：
文献类型：
--
作者：
Dustin Tran;M. Hoffman;R. Saurous;E. Brevdo;K. Murphy;D. Blei
我们建议爱德华（Edward），这是一种完整的概率编程语言。爱德华定义了两个组成表示---随机变量和推理。通过将推论视为一流的公民，在建模方面，我们表明概率编程可以像传统的深度学习一样灵活且计算上有效。为了灵活性，爱德华（Edward）可以轻松使用各种可合并的推理方法拟合相同的模型，从点估计到变分推断，再到MCMC。此外，爱德华可以将建模表示作为推理的一部分重复使用，从而促进了丰富的变分模型和生成对抗网络的设计。为了提高效率，爱德华被整合到张力流中，对现有概率系统提供了显着的加速。例如，我们在基准的逻辑回归任务上显示，爱德华至少比Stan快35倍，比PYMC3快6倍。此外，爱德华（Edward）不带有运行时开销：它的速度与手写的TensorFlow一样快。
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.