1 April 2011, Computational Social Choice Seminar, Amy Greenwald
In this talk, I will describe QuickRank, an efficient algorithm for ranking individuals in a society, given a network that encodes their relationships, assuming that network possesses an accompanying hierarchical structure: e.g., a citation index together with a topic hierarchy. The QuickRank design is motivated by both computational concerns, and intuitive goals including the ``peer-review'' principle and a hypothesis due to Bonacich. Together, these objectives lead to a recursive ranking algorithm which is scalable, parallelizable, and easily updateable. It is also seemingly more resistant to link-spamming than other popular ranking algorithms. Joint work with John Wicks.