BriefGPT.xyz
Jun, 2020
基于神经网络的逐点依存关系估计方法
Neural Methods for Point-wise Dependency Estimation
HTML
PDF
Yao-Hung Hubert Tsai, Han Zhao, Makoto Yamada, Louis-Philippe Morency, Ruslan Salakhutdinov
TL;DR
本文提出了两种方法（Probabilistic Classifier和Density-Ratio Fitting）来解决MI估计中存在的方差大的问题，这两种方法可以自然地获得点对点的依赖信息，为自监督学习和跨模态检索任务提供了有效方法。
Abstract
Since its inception, the
neural estimation
of
mutual information
(MI) has demonstrated the empirical success of modeling expected dependency between high-dimensional random variables. However, MI is an aggregate
→