BriefGPT.xyz
Jun, 2018
超越重新参数化技巧的路径导数
Pathwise Derivatives Beyond the Reparameterization Trick
HTML
PDF
Martin Jankowiak, Fritz Obermeyer
TL;DR
通过重参数化技巧计算的梯度与最优输运形式中的传输方程解直接对应。我们使用这种视角来为直接不易受到重参数化技巧影响的概率分布(伽马、贝塔和Dirichlet)计算(近似)路径梯度。
Abstract
We observe that gradients computed via the
reparameterization trick
are in direct correspondence with solutions of the transport equation in the formalism of
optimal transport
. We use this perspective to compute
→