BriefGPT.xyz
Nov, 2018
学习深度核函数用于指数族密度
Learning deep kernels for exponential family densities
HTML
PDF
Li Wenliang, Danica Sutherland, Heiko Strathmann, Arthur Gretton
TL;DR
提供了一种通过学习深度网络参数化的核函数来建模复杂结构的密度模型方法,相较于利用最大似然拟合的深度密度模型,虽然前者可能会得到更高似然度,但后者提供了更好描述分布形状的对数密度梯度得到更好的估计,二者有不同的优缺点。
Abstract
The
kernel exponential family
is a rich class of distributions,which can be fit efficiently and with statistical guarantees by
score matching
. Being required to choose a priori a simple kernel such as the Gaussia
→