BriefGPT.xyz
Feb, 2021
VAE近似误差: ELBO与指数族
VAE Approximation Error: ELBO and Conditional Independence
HTML
PDF
Dmitrij Schlesinger, Alexander Shekhovtsov, Boris Flach
TL;DR
本研究分析了 Variational Autoencoders 的近似误差，探讨了该误差的多种可能性并找到了其一致子集。重要的是，此子集无法通过考虑更深的编码器/解码器网络进行扩大，也无法降低相应误差。
Abstract
The importance of
variational autoencoders
reaches far beyond standalone generative models -- the approach is also used for learning
latent representations
and can be generalized to
→