BriefGPT.xyz
Oct, 2022
通过反向传播学习离散有向无环图
Learning Discrete Directed Acyclic Graphs via Backpropagation
HTML
PDF
Andrew J. Wren, Pasquale Minervini, Luca Franceschi, Valentina Zantedeschi
TL;DR
本文提出DAG-DB框架，采用离散反向传播技术从数据中学习有向无环图，其使用了内在概率分布采样二元邻接矩阵，然后学习分布的参数。
Abstract
Recently
continuous relaxations
have been proposed in order to learn
directed acyclic graphs
(DAGs) from data by
backpropagation
, instead
→