Note to moderator: Read the answerline carefully.David Kempe, Jon Kleinberg, and Éva Tardos proved guarantees for algorithms on these data structures that approximate answers to a problem posed by Pedro Domingos and Matt Richardson. Messages on these data structures are updated using either a sum-product rule or a max-product rule in belief propagation algorithms. An algorithm on these data structures uses teleportation operations to avoid spider traps; that algorithm operates on one of these data structures with a bow tie-like structure. In spectral clustering, k-means clustering is performed on the eigenvectors of one of these data structures’ Laplacian matrix. PageRank operates on one of these data structures built from web pages. A giant connected component is contained in one of these data structures that represents the World Wide Web. For 10 points, networks are represented with what data structures that contain nodes and edges? ■END■
ANSWER: graphs [accept undirected graphs, directed graphs, graphical models, social graphs, social network graphs, or webgraphs; accept trees; prompt on social networks or Bayesian networks until “networks” is read by asking “what data structures are used to represent such networks?”; prompt on Markov random fields after “sum-product rule” is read by asking “what data structures are used to represent Markov random fields?”; prompt on adjacent lists or adjacency matrixes or adjacency matrices by asking “those are representations of what data structures?”; reject “neural networks,” “matrix,” “matrices,” or “lists”] (The first sentence is about algorithms for influence spreading through a social network.)
<Other Science>
= Average correct buzz position