BriefGPT.xyz
Jul, 2019
使用宇宙、群组和任务特征学习的多任务网络
Multi-Task Networks With Universe, Group, and Task Feature Learning
HTML
PDF
Shiva Pentyala, Mengwen Liu, Markus Dreyer
TL;DR
通过自然的相关任务分组，结合多任务学习方法，将监督信息编码到模型中，探索并实现了两种神经网络结构，分别在不同层级学习不同的特征空间，并在自然语言理解中取得了显著的性能提升。
Abstract
We present methods for
multi-task learning
that take advantage of natural groupings of related tasks.
task groups
may be defined along known properties of the tasks, such as task domain or language. Such
→