Machine Learning Methods for Network Visualization and Analysis

Abstract

Networks are commonly used for representing complex systems, such as interactions between proteins, data communications between computers, and relationships between people. Recently, machine learning methods have achieved impressive performance for a variety of data analysis tasks. This talk presents machine learning methods for visualizing and analyzing network data. First, I present a method designed to quickly visualize a network in a chosen layout based on computing its structural similarity with a database of pre-laid-out networks using graph kernels. Second, I introduce a technique to systematically generate visualizations of a network in diverse layouts using deep generative models. In summary, this talk presents new approaches to depicting networks where users can effortlessly obtain effective visualization of a network without expert knowledge.