arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1810.04472cs.LGcs.AIstat.ML
收藏

无监督自适应的自集成域混淆算法

Domain Confusion with Self Ensembling for Unsupervised Adaptation

Jiawei Wang, Zhaoshui He, Chengjian Feng, Zhouping Zhu, Qinzhuang Lin, Jun Lv, Shengli Xie

摘要:

在机器学习中,数据汇集和标注是一个非常耗时的过程,特别是对于大规模的问题。解决这一问题的一种常用方法是将知识从一个相关的标记域转移到一个目标域。有两种流行的方法来实现这一目标:对抗性学习和自我训练。本文首先分析了对抗式学习过程中的训练不稳定性问题和错误混淆问题。然后,受领域混淆和自集成方法的启发,我们提出了一个组合模型来学习特征和类的联合不变表示,即领域混淆与自集成(DCSE)。实验结果表明,该方法在各种无监督领域自适应测试中均能获得比经验自适应方法更好的性能。

英文摘要:

Data collection and annotation are time-consuming in machine learning, expecially for large scale problem. A common approach for this problem is to transfer knowledge from a related labeled domain to a target one. There are two popular ways to achieve this goal: adversarial learning and self training. In this article, we first analyze the training unstablity problem and the mistaken confusion issue in adversarial learning process. Then, inspired by domain confusion and self-ensembling methods, we propose a combined model to learn feature and class jointly invariant representation, namely Domain Confusion with Self Ensembling (DCSE). The experiments verified that our proposed approach can offer better performance than empirical art in a variety of unsupervised domain adaptation benchmarks.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑