Active Learning for Hidden Attributes in Networks - Statistics > Machine LearningReport as inadecuate




Active Learning for Hidden Attributes in Networks - Statistics > Machine Learning - Download this document for free, or read online. Document in PDF available to download.

Abstract: In many networks, vertices have hidden attributes, or types, that arecorrelated with the networks topology. If the topology is known but theseattributes are not, and if learning the attributes is costly, we need a methodfor choosing which vertex to query in order to learn as much as possible aboutthe attributes of the other vertices. We assume the network is generated by astochastic block model, but we make no assumptions about its assortativity ordisassortativity. We choose which vertex to query using two methods: 1maximizing the mutual information between its attributes and those of theothers a well-known approach in active learning and 2 maximizing the averageagreement between two independent samples of the conditional Gibbsdistribution. Experimental results show that both these methods do much betterthan simple heuristics. They also consistently identify certain vertices asimportant by querying them early on.



Author: Xiaoran Yan, Yaojia Zhu, Jean-Baptiste Rouquier, Cristopher Moore

Source: https://arxiv.org/







Related documents