BriefGPT.xyz
Mar, 2021
MetaAlign: 无监督领域自适应中的领域对齐和分类协调
MetaAlign: Coordinating Domain Alignment and Classification for Unsupervised Domain Adaptation
HTML
PDF
Guoqiang Wei, Cuiling Lan, Wenjun Zeng, Zhibo Chen
TL;DR
本文介绍了一个基于元学习的有效优化策略MetaAlign,以协调域对齐任务和分类任务的目标函数的不一致性,实现在元训练和元测试中同时优化域对齐和分类任务的最大化,从而提高了无监督领域自适应问题的性能。
Abstract
For
unsupervised domain adaptation
(UDA), to alleviate the effect of domain shift, many approaches align the source and target domains in the feature space by adversarial learning or by explicitly aligning their statistics. However, the
→