Our idea is to take some of the ideas from the uncertainty quantification and apply them to the PageRank equation. From an uncertainty quantification perspective, many computational models have a set of parameters that are fit from data or chosen to reproduce some desired behavior. However, these choices may be inaccurate and the goal of uncertainty quantification is to determine a reasonable error bound for the solution when random variables substitute for the deterministic parameters.
The PageRank model has three parameters, a web graph, a teleportation vector, and a teleportation coefficient. We began our investigation by looking at replacing the teleportation coefficient with a random variable distributed according to a Beta distribution.
On this website, we've summarized our current work on this topic.
Papers and Publications
- (with Paul Constantine), Using polynomial chaos to compute the influence of multiple random surfers in the PageRank model, Workshop on Algorithms for the Web Graph, 2007, pp. 82--95. (Conference Publication)
- What to expect with randomized PageRank and how polynomial chaos simplifies it, Seminar on Linear Algebra and Optimization, 31 October 2007. (Presentation)
- (with Paul Constantine), Really Random PageRank, Workshop on Algorithms for the Web Graph, 2007, 11 December 2007. (Presentation)
Codes
- Coming soon, I promise! We're trying to make them as nice as possible.