Bayesian attributed network sampling

Sampling;Data Mining;Social and Information Networks;Bayesian Analysis

Kumar, Ankit ; Sundaram ; Hari

关键词: Sampling;    Data Mining;    Social and Information Networks;    Bayesian Analysis;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/108049/KUMAR-THESIS-2020.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

We address the problem of sampling in attributed networks. While uniform sampling is a task independent sampling method, in real-world, this is often difficult to implement as it requires random access to all the nodes of graph. Link tracing sampling methods such as random Walk, expansion sampling overcome this problem, however they do not utilize the information provided by attributes of the nodes and just use the topology of the graph. We propose a network sampling method which is task independent and utilizes the node attributes. Our approach is based on introducing maximum unfamiliarity in each sampling step and it uses Bayesian approach to asses the familiarity of neighboring nodes with respect to the current sample.