BriefGPT.xyz
Oct, 2018
深度模型变分贝叶斯的良好初始化
Good Initializations of Variational Bayes for Deep Models
HTML
PDF
Simone Rossi, Pietro Michiardi, Maurizio Filippone
TL;DR
本文提出了一种基于贝叶斯线性模型的新型逐层初始化策略，用于解决随机变分推断的初始化问题，并在回归、分类等任务中进行了广泛验证，与启发式初始化相比，能够实现更快更好的收敛。
Abstract
stochastic variational inference
is an established way to carry out approximate Bayesian inference for
deep models
. While there have been effective proposals for good initializations for loss minimization in deep
→