This website requires JavaScript.

Unsupervised Instance and Subnetwork Selection for Network Data

Lin ZhangNicholas MoskwaMelinda LarsenPetko Bogdanov
Dec 2022
摘要
Unlike tabular data, features in network data are interconnected within adomain-specific graph. Examples of this setting include gene expressionoverlaid on a protein interaction network (PPI) and user opinions in a socialnetwork. Network data is typically high-dimensional (large number of nodes) andoften contains outlier snapshot instances and noise. In addition, it is oftennon-trivial and time-consuming to annotate instances with global labels (e.g.,disease or normal). How can we jointly select discriminative subnetworks andrepresentative instances for network data without supervision? We address thesechallenges within an unsupervised framework for joint subnetwork and instanceselection in network data, called UISS, via a convex self-representationobjective. Given an unlabeled network dataset, UISS identifies representativeinstances while ignoring outliers. It outperforms state-of-the-art baselines onboth discriminative subnetwork selection and representative instance selection,achieving up to 10% accuracy improvement on all real-world data sets we use forevaluation. When employed for exploratory analysis in RNA-seq network samplesfrom multiple studies it produces interpretable and informative summaries.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答