BriefGPT.xyz
Feb, 2016
辅助深度生成模型
Auxiliary Deep Generative Models
HTML
PDF
Lars Maaløe, Casper Kaae Sønderby, Søren Kaae Sønderby, Ole Winther
TL;DR
通过引入辅助变量(auxiliary variables),我们扩展了深度生成模型,用于提高变分近似方法,这让变分分布更有表现力,具有更快的收敛速度和更好的结果,从而实现了半监督学习的最先进性能,适用于MNIST、SVHN和NORB数据集。
Abstract
deep generative models
parameterized by
neural networks
have recently achieved state-of-the-art performance in unsupervised and
semi-supervised l
→