arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1807.04489stat.MLcs.ITcs.LGmath.ITstat.CO
收藏

复杂模型变分推断的快速简单自然梯度下降法

Fast yet Simple Natural-Gradient Descent for Variational Inference in Complex Models

RIKEN Center for Advanced Intelligence Project(理化学研究所先进智能项目中心)

Mohammad Emtiyaz Khan, Didrik Nielsen

摘要:

贝叶斯推断在推进机器学习方面发挥着重要作用,但在应用于深度神经网络等复杂模型时面临着计算挑战。变分推断通过将贝叶斯推断公式化为优化问题并使用基于梯度的优化来解决它来规避这些挑战。在本文中,我们主张自然梯度的方法,不像他们的梯度为基础的同行,可以提高收敛利用信息几何的解决方案。我们展示了如何通过使用与指数族分布相关的对偶来获得快速而简单的自然梯度更新。这些方法的一个有吸引力的特点是,通过使用自然梯度,它们能够为单个模型组件提取精确的局部近似。我们总结了贝叶斯深度学习的最新结果,显示了自然梯度方法相对于梯度方法的优越性。

英文摘要:

Bayesian inference plays an important role in advancing machine learning, but faces computational challenges when applied to complex models such as deep neural networks. Variational inference circumvents these challenges by formulating Bayesian inference as an optimization problem and solving it using gradient-based optimization. In this paper, we argue in favor of natural-gradient approaches which, unlike their gradient-based counterparts, can improve convergence by exploiting the information geometry of the solutions. We show how to derive fast yet simple natural-gradient updates by using a duality associated with exponential-family distributions. An attractive feature of these methods is that, by using natural-gradients, they are able to extract accurate local approximations for individual model components. We summarize recent results for Bayesian deep learning showing the superiority of natural-gradient approaches over their gradient counterparts.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑