arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2103.13145cs.LG
收藏

使用特权信息的无监督协作学习

Unsupervised collaborative learning using privileged information

Université Sorbonne Paris Nord(巴黎北索邦大学);CNRS(法国国家科学研究中心);Institut Galilée(加利莱研究所)

Yohan Foucade, Younès Bennani

摘要:

在协作聚类框架中,希望通过组合多个聚类解决方案,每个解决方案都有自己的偏见和不完善之处,从而得到一个更好的整体解决方案。目标是每个本地计算,很可能应用于不同的数据集,从其他合作者所做的工作中受益。这篇文章致力于基于学习使用已标记信息范例的协作聚类。局部算法在每个观察的级别上对传入信息进行加权,这取决于该观察的分类的置信水平。我们的算法和最先进的实现之间的比较表明,使用所提出的方法的协作过程的改进。

英文摘要:

In the collaborative clustering framework, the hope is that by combining several clustering solutions, each one with its own bias and imperfections, one will get a better overall solution. The goal is that each local computation, quite possibly applied to distinct data sets, benefits from the work done by the other collaborators. This article is dedicated to collaborative clustering based on the Learning Using Privileged Information paradigm. Local algorithms weight incoming information at the level of each observation, depending on the confidence level of the classification of that observation. A comparison between our algorithm and state of the art implementations shows improvement of the collaboration process using the proposed approach.

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。

↑