NetGist: Learning to Generate Task-Based Network Summaries

摘要: Given a network, can we visualize it for any given task, highlighting the important characteristics? Networks are widespread, and hence summarizing visualizing them is of primary interest many applications such as viral marketing, extracting communities immunization. Summaries help in solving new problems visualization, sense-making, other goals. However, most prior work focuses on generic structural summarization techniques or developing specific algorithms tasks. This both tedious challenging. As result, several popular tasks, there do not exist readymade methods. In this paper, explore promising alternative approach instead. We propose NetGist, framework which automatically learns how to generate summary task network. addition generating required summary, also allows us reuse learned process similar networks. formulate novel task-based graph problem leverage reinforcement learning design flexible our solution. Via extensive experiments, show that NetGist robustly effectively meaningful summaries, helps solve challenging problems, aids complex sense-making