This project was completed in December of 2013. The team brought together machine learning and information theory researchers to estimate models of graph data from network samples. For example, computing page rank for search engines, or peer influence in social networks both rely on an estimate of the underlying network model. However, in many cases the underlying graph is unknown, and must be estimated through sampling. By performing random walks several questions arise in terms of inferring the accuracy of the structure of the underlying graph, and what tradeoffs one must consider between compression rates of the data and the fidelity of the graph being reproduced. This project was active from September 1, 2012 - December 1, 2013.

A presentation of this project is avaliable here.