BriefGPT.xyz
Jul, 2018
复杂模型变分推断的快速简单的自然梯度下降
Fast yet Simple Natural-Gradient Descent for Variational Inference in Complex Models
HTML
PDF
Mohammad Emtiyaz Khan, Didrik Nielsen
TL;DR
本文研究使用自然梯度方法解决贝叶斯深度学习中的复杂模型的计算挑战，并表明这种方法在提高收敛性方面有优势。
Abstract
bayesian inference
plays an important role in advancing
machine learning
, but faces computational challenges when applied to complex models such as
→