arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2105.04051cs.LGstat.ML
收藏

从多个目标偏移源聚合

Aggregating From Multiple Target-Shifted Sources

Université Laval(拉瓦尔大学);Guangdong University of Technology(广东工业大学);Western University(韦仕敦大学);Mila(米拉人工智能研究所);Vector Institute(向量研究所)

Changjian Shui, Zijian Li, Jiaqi Li, Christian Gagné, Charles Ling, Boyu Wang

摘要:

多源领域自适应的目的是利用多个任务的知识来预测相关的目标领域。因此,一个关键的方面是根据它们的关系适当地组合不同的来源。在本文中,我们分析了聚合源域不同的标签分布,最新的源选择方法失败的问题。我们提出的算法在两个关键方面不同于以前的方法:该模型主要通过语义条件分布的相似性而不是边缘分布来聚合多个源;该模型提出了一个\emph{unified}框架来选择三个流行场景的相关源,即,在目标域上具有有限标记的领域自适应、无监督领域自适应和标记部分无监督域自适应。我们通过大量的实验评估所提出的方法。实证结果显着优于基线。

英文摘要:

Multi-source domain adaptation aims at leveraging the knowledge from multiple tasks for predicting a related target domain. Hence, a crucial aspect is to properly combine different sources based on their relations. In this paper, we analyzed the problem for aggregating source domains with different label distributions, where most recent source selection approaches fail. Our proposed algorithm differs from previous approaches in two key ways: the model aggregates multiple sources mainly through the similarity of semantic conditional distribution rather than marginal distribution; the model proposes a \emph{unified} framework to select relevant sources for three popular scenarios, i.e., domain adaptation with limited label on target domain, unsupervised domain adaptation and label partial unsupervised domain adaption. We evaluate the proposed method through extensive experiments. The empirical results significantly outperform the baselines.

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。

↑